perf(db): disable prefault_write to avoid mincore() syscall overhead

In WRITEMAP mode, MDBX by default calls mincore() before each page write to check if
the page is resident in memory. Profiling shows this accounts for ~20-22% of
persistence time, yet pages are usually already warm from Engine Task reads.

Disabling prefault_write skips the mincore() syscall and lets the kernel handle
page faults directly. Since pages are likely resident, this trades potential
page faults (which rarely occur due to cache warmth) for syscall overhead
elimination.

Amp-Thread-ID: https://ampcode.com/threads/T-019c233a-3fa9-733b-8985-991c393eb610
Co-authored-by: Amp <amp@ampcode.com>

.github/scripts/hive/build_simulators.sh

@@ -38,6 +38,6 @@ for pid in "${saving_pids[@]}"; do
 done
 
 # Make sure we don't rebuild images on the CI jobs
-git apply ../.github/assets/hive/no_sim_build.diff
+git apply ../.github/scripts/hive/no_sim_build.diff
 go build .
 mv ./hive ../hive_assets/

.github/workflows/e2e.yml

@@ -44,3 +44,24 @@ jobs:
             --exclude 'op-reth' \
             --exclude 'reth' \
             -E 'binary(e2e_testsuite)'
+
+  rocksdb:
+    name: e2e-rocksdb
+    runs-on: depot-ubuntu-latest-4
+    env:
+      RUST_BACKTRACE: 1
+    timeout-minutes: 60
+    steps:
+      - uses: actions/checkout@v6
+      - uses: dtolnay/rust-toolchain@stable
+      - uses: mozilla-actions/sccache-action@v0.0.9
+      - uses: taiki-e/install-action@nextest
+      - uses: Swatinem/rust-cache@v2
+        with:
+          cache-on-failure: true
+      - name: Run RocksDB e2e tests
+        run: |
+          cargo nextest run \
+            --locked --features "edge" \
+            -p reth-e2e-test-utils \
+            -E 'binary(rocksdb)'

.github/workflows/hive.yml

@@ -58,11 +58,11 @@ jobs:
         uses: actions/cache@v5
         with:
           path: ./hive_assets
-          key: hive-assets-${{ steps.hive-commit.outputs.hash }}-${{ hashFiles('.github/assets/hive/build_simulators.sh') }}
+          key: hive-assets-${{ steps.hive-commit.outputs.hash }}-${{ hashFiles('.github/scripts/hive/build_simulators.sh') }}
 
       - name: Build hive assets
         if: steps.cache-hive.outputs.cache-hit != 'true'
-        run: .github/assets/hive/build_simulators.sh
+        run: .github/scripts/hive/build_simulators.sh
 
       - name: Load cached Docker images
         if: steps.cache-hive.outputs.cache-hit == 'true'
@@ -213,7 +213,7 @@ jobs:
           path: /tmp
 
       - name: Load Docker images
-        run: .github/assets/hive/load_images.sh
+        run: .github/scripts/hive/load_images.sh
 
       - name: Move hive binary
         run: |
@@ -241,11 +241,11 @@ jobs:
             FILTER="/"
           fi
           echo "filter: $FILTER"
-          .github/assets/hive/run_simulator.sh "${{ matrix.scenario.sim }}" "$FILTER"
+          .github/scripts/hive/run_simulator.sh "${{ matrix.scenario.sim }}" "$FILTER"
 
       - name: Parse hive output
         run: |
-          find hivetests/workspace/logs -type f -name "*.json" ! -name "hive.json" | xargs -I {} python .github/assets/hive/parse.py {} --exclusion .github/assets/hive/expected_failures.yaml --ignored .github/assets/hive/ignored_tests.yaml
+          find hivetests/workspace/logs -type f -name "*.json" ! -name "hive.json" | xargs -I {} python .github/scripts/hive/parse.py {} --exclusion .github/scripts/hive/expected_failures.yaml --ignored .github/scripts/hive/ignored_tests.yaml
 
       - name: Print simulator output
         if: ${{ failure() }}

.github/workflows/integration.yml

@@ -22,7 +22,7 @@ concurrency:
 
 jobs:
   test:
-    name: test / ${{ matrix.network }}
+    name: test / ${{ matrix.network }} / ${{ matrix.storage }}
     if: github.event_name != 'schedule'
     runs-on: depot-ubuntu-latest-4
     env:
@@ -30,13 +30,17 @@ jobs:
     strategy:
       matrix:
         network: ["ethereum", "optimism"]
+        storage: ["stable", "edge"]
+        exclude:
+          - network: optimism
+            storage: edge
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v6
       - uses: rui314/setup-mold@v1
       - uses: dtolnay/rust-toolchain@stable
       - name: Install Geth
-        run: .github/assets/install_geth.sh
+        run: .github/scripts/install_geth.sh
       - uses: taiki-e/install-action@nextest
       - uses: mozilla-actions/sccache-action@v0.0.9
       - uses: Swatinem/rust-cache@v2
@@ -46,7 +50,7 @@ jobs:
         name: Run tests
         run: |
           cargo nextest run \
-            --locked --features "asm-keccak ${{ matrix.network }}" \
+            --locked --features "asm-keccak ${{ matrix.network }} ${{ matrix.storage == 'edge' && 'edge' || '' }}" \
             --workspace --exclude ef-tests \
             -E "kind(test) and not binary(e2e_testsuite)"
       - if: matrix.network == 'optimism'

.github/workflows/label-pr.yml

@@ -19,5 +19,5 @@ jobs:
         uses: actions/github-script@v8
         with:
           script: |
-            const label_pr = require('./.github/assets/label_pr.js')
+            const label_pr = require('./.github/scripts/label_pr.js')
             await label_pr({github, context})

.github/workflows/lint.yml

@@ -76,7 +76,7 @@ jobs:
       - name: Run Wasm checks
         run: |
           sudo apt update && sudo apt install gcc-multilib
-          .github/assets/check_wasm.sh
+          .github/scripts/check_wasm.sh
 
   riscv:
     runs-on: depot-ubuntu-latest
@@ -94,7 +94,7 @@ jobs:
           cache-on-failure: true
       - uses: dcarbone/install-jq-action@v3
       - name: Run RISC-V checks
-        run: .github/assets/check_rv32imac.sh
+        run: .github/scripts/check_rv32imac.sh
 
   crate-checks:
     name: crate-checks (${{ matrix.partition }}/${{ matrix.total_partitions }})

.github/workflows/prepare-reth.yml

@@ -43,7 +43,7 @@ jobs:
         uses: docker/build-push-action@v6
         with:
           context: .
-          file: .github/assets/hive/Dockerfile
+          file: .github/scripts/hive/Dockerfile
           tags: ${{ inputs.image_tag }}
           outputs: type=docker,dest=./artifacts/reth_image.tar
           build-args: |

CLAUDE.md

@@ -38,7 +38,7 @@ Reth is a high-performance Ethereum execution client written in Rust, focusing o
 
 2. **Linting**: Run clippy with all features
    ```bash
-   RUSTFLAGS="-D warnings" cargo +nightly clippy --workspace --lib --examples --tests --benches --all-features --locked
+   cargo +nightly clippy --workspace --lib --examples --tests --benches --all-features 
    ```
 
 3. **Testing**: Use nextest for faster test execution
@@ -169,12 +169,11 @@ Based on PR patterns, avoid:
 Before submitting changes, ensure:
 
 1. **Format Check**: `cargo +nightly fmt --all --check`
-2. **Clippy**: No warnings with `RUSTFLAGS="-D warnings"`
+2. **Clippy**: No warnings
 3. **Tests Pass**: All unit and integration tests
 4. **Documentation**: Update relevant docs and add doc comments with `cargo docs --document-private-items`
 5. **Commit Messages**: Follow conventional format (feat:, fix:, chore:, etc.)
 
-
 ### Opening PRs against <https://github.com/paradigmxyz/reth>
 
 Label PRs appropriately, first check the available labels and then apply the relevant ones:
@@ -349,10 +348,10 @@ Let's say you want to fix a bug where external IP resolution fails on startup:
    }
    ```
 
-5. **Run checks**:
+5. **Run checks** (IMPORTANT!):
    ```bash
    cargo +nightly fmt --all
-   cargo clippy --all-features
+   cargo clippy --workspace --all-features # Make sure WHOLE WORKSPACE compiles!
    cargo test -p reth-discv4
    ```
 
@@ -374,7 +373,7 @@ Let's say you want to fix a bug where external IP resolution fails on startup:
 cargo +nightly fmt --all
 
 # Run lints
-RUSTFLAGS="-D warnings" cargo +nightly clippy --workspace --all-features --locked
+cargo +nightly clippy --workspace --all-features
 
 # Run tests
 cargo nextest run --workspace

Cargo.toml

@@ -77,6 +77,10 @@ members = [
     "crates/optimism/cli",
     "crates/optimism/consensus",
     "crates/optimism/evm/",
+    "crates/optimism/examples/custom-node",
+    "crates/optimism/examples/engine-api-access",
+    "crates/optimism/examples/exex-hello-world",
+    "crates/optimism/examples/op-db-access",
     "crates/optimism/flashblocks/",
     "crates/optimism/hardforks/",
     "crates/optimism/node/",
@@ -145,7 +149,6 @@ members = [
     "examples/beacon-api-sse/",
     "examples/bsc-p2p",
     "examples/custom-dev-node/",
-    "examples/custom-node/",
     "examples/custom-engine-types/",
     "examples/custom-evm/",
     "examples/custom-hardforks/",
@@ -154,10 +157,7 @@ members = [
     "examples/custom-payload-builder/",
     "examples/custom-rlpx-subprotocol",
     "examples/custom-rpc-middleware",
-    "examples/custom-node",
     "examples/db-access",
-    "examples/engine-api-access",
-    "examples/exex-hello-world",
     "examples/exex-subscription",
     "examples/exex-test",
     "examples/full-contract-state",
@@ -168,7 +168,6 @@ members = [
     "examples/node-builder-api/",
     "examples/node-custom-rpc/",
     "examples/node-event-hooks/",
-    "examples/op-db-access/",
     "examples/polygon-p2p/",
     "examples/rpc-db/",
     "examples/precompile-cache/",
@@ -484,15 +483,15 @@ op-revm = { version = "15.0.0", default-features = false }
 revm-inspectors = "0.34.1"
 
 # eth
+alloy-dyn-abi = "1.5.4"
+alloy-primitives = { version = "1.5.4", default-features = false, features = ["map-foldhash"] }
+alloy-sol-types = { version = "1.5.4", default-features = false }
+
 alloy-chains = { version = "0.2.5", default-features = false }
-alloy-dyn-abi = "1.5.2"
 alloy-eip2124 = { version = "0.2.0", default-features = false }
 alloy-eip7928 = { version = "0.3.0", default-features = false }
 alloy-evm = { version = "0.27.0", default-features = false }
-alloy-primitives = { version = "1.5.0", default-features = false, features = ["map-foldhash"] }
 alloy-rlp = { version = "0.3.10", default-features = false, features = ["core-net"] }
-alloy-sol-macro = "1.5.0"
-alloy-sol-types = { version = "1.5.0", default-features = false }
 alloy-trie = { version = "0.9.1", default-features = false }
 
 alloy-hardforks = "0.4.5"
@@ -561,7 +560,7 @@ humantime-serde = "1.1"
 itertools = { version = "0.14", default-features = false }
 linked_hash_set = "0.1"
 lz4 = "1.28.1"
-modular-bitfield = "0.11.2"
+modular-bitfield = "0.13.1"
 notify = { version = "8.0.0", default-features = false, features = ["macos_fsevent"] }
 nybbles = { version = "0.4.2", default-features = false }
 once_cell = { version = "1.19", default-features = false, features = ["critical-section"] }
@@ -590,13 +589,13 @@ zstd = "0.13"
 byteorder = "1"
 fixed-cache = { version = "0.1.7", features = ["stats"] }
 moka = "0.12"
-tar-no-std = { version = "0.3.2", default-features = false }
-miniz_oxide = { version = "0.8.4", default-features = false }
+tar-no-std = { version = "0.4.2", default-features = false }
+miniz_oxide = { version = "0.9.0", default-features = false }
 chrono = "0.4.41"
 
 # metrics
 metrics = "0.24.0"
-metrics-derive = "0.1"
+metrics-derive = "0.1.1"
 metrics-exporter-prometheus = { version = "0.18.0", default-features = false }
 metrics-process = "2.1.0"
 metrics-util = { default-features = false, version = "0.20.0" }
@@ -608,7 +607,7 @@ quote = "1.0"
 # tokio
 tokio = { version = "1.44.2", default-features = false }
 tokio-stream = "0.1.11"
-tokio-tungstenite = "0.26.2"
+tokio-tungstenite = "0.28.0"
 tokio-util = { version = "0.7.4", features = ["codec"] }
 
 # async
@@ -621,7 +620,7 @@ futures-util = { version = "0.3", default-features = false }
 hyper = "1.3"
 hyper-util = "0.1.5"
 pin-project = "1.0.12"
-reqwest = { version = "0.12", default-features = false }
+reqwest = { version = "0.12", default-features = false, features = ["rustls-tls", "rustls-tls-native-roots", "stream"] }
 tracing-futures = "0.2"
 tower = "0.5"
 tower-http = "0.6"
@@ -641,7 +640,6 @@ jsonrpsee-types = "0.26.0"
 http = "1.0"
 http-body = "1.0"
 http-body-util = "0.1.2"
-jsonwebtoken = "9"
 proptest-arbitrary-interop = "0.1.0"
 
 # crypto
@@ -655,7 +653,7 @@ rand_08 = { package = "rand", version = "0.8" }
 c-kzg = "2.1.5"
 
 # config
-toml = "0.8"
+toml = "0.9"
 
 # rocksdb
 rocksdb = { version = "0.24" }
@@ -671,19 +669,19 @@ tracing-opentelemetry = "0.32"
 # misc-testing
 arbitrary = "1.3"
 assert_matches = "1.5.0"
-criterion = { package = "codspeed-criterion-compat", version = "2.7" }
+criterion = { package = "codspeed-criterion-compat", version = "4.3" }
 insta = "1.41"
 proptest = "1.7"
-proptest-derive = "0.5"
+proptest-derive = "0.7"
 similar-asserts = { version = "1.5.0", features = ["serde"] }
 tempfile = "3.20"
 test-fuzz = "7"
-rstest = "0.24.0"
+rstest = "0.26.1"
 test-case = "3"
 
 # ssz encoding
-ethereum_ssz = "0.9.0"
-ethereum_ssz_derive = "0.9.0"
+ethereum_ssz = "0.10.1"
+ethereum_ssz_derive = "0.10.1"
 
 # allocators
 jemalloc_pprof = { version = "0.8", default-features = false }
@@ -695,14 +693,14 @@ snmalloc-rs = { version = "0.3.7", features = ["build_cc"] }
 aes = "0.8.1"
 ahash = "0.8"
 anyhow = "1.0"
-bindgen = { version = "0.71", default-features = false }
-block-padding = "0.3.2"
+bindgen = { version = "0.72", default-features = false }
+block-padding = "0.3"
 cc = "1.2.15"
 cipher = "0.4.3"
 comfy-table = "7.0"
 concat-kdf = "0.1.0"
 crossbeam-channel = "0.5.13"
-crossterm = "0.28.0"
+crossterm = "0.29.0"
 csv = "1.3.0"
 ctrlc = "3.4"
 ctr = "0.9.2"
@@ -715,7 +713,7 @@ hmac = "0.12.1"
 human_bytes = "0.4.1"
 indexmap = "2"
 interprocess = "2.2.0"
-lz4_flex = { version = "0.11", default-features = false }
+lz4_flex = { version = "0.12", default-features = false }
 memmap2 = "0.9.4"
 mev-share-sse = { version = "0.5.0", default-features = false }
 num-traits = "0.2.15"
@@ -723,17 +721,17 @@ page_size = "0.6.0"
 parity-scale-codec = "3.2.1"
 plain_hasher = "0.2"
 pretty_assertions = "1.4"
-ratatui = { version = "0.29", default-features = false }
-ringbuffer = "0.15.0"
+ratatui = { version = "0.30", default-features = false }
+ringbuffer = "0.16.0"
 rmp-serde = "1.3"
-roaring = "0.10.2"
+roaring = "0.11.3"
 rolling-file = "0.2.0"
 sha3 = "0.10.5"
 snap = "1.1.1"
-socket2 = { version = "0.5", default-features = false }
-sysinfo = { version = "0.33", default-features = false }
+socket2 = { version = "0.6", default-features = false }
+sysinfo = { version = "0.38", default-features = false }
 tracing-journald = "0.3"
-tracing-logfmt = "0.3.3"
+tracing-logfmt = "=0.3.5"
 tracing-samply = "0.1"
 tracing-subscriber = { version = "0.3", default-features = false }
 tracing-tracy = "0.11"

Dockerfile.depot

@@ -5,7 +5,7 @@
 #   reth:    --build-arg BINARY=reth
 #   op-reth: --build-arg BINARY=op-reth --build-arg MANIFEST_PATH=crates/optimism/bin
 
-FROM lukemathwalker/cargo-chef:latest-rust-1 AS chef
+FROM rust:1 AS builder
 WORKDIR /app
 
 LABEL org.opencontainers.image.source=https://github.com/paradigmxyz/reth
@@ -19,14 +19,6 @@ ENV RUSTC_WRAPPER=sccache
 ENV SCCACHE_DIR=/sccache
 ENV SCCACHE_WEBDAV_ENDPOINT=https://cache.depot.dev
 
-# Builds a cargo-chef plan
-FROM chef AS planner
-COPY --exclude=.git . .
-RUN cargo chef prepare --recipe-path recipe.json
-
-FROM chef AS builder
-COPY --from=planner /app/recipe.json recipe.json
-
 # Binary to build (reth or op-reth)
 ARG BINARY=reth
 
@@ -53,13 +45,6 @@ ENV VERGEN_GIT_SHA=$VERGEN_GIT_SHA
 ENV VERGEN_GIT_DESCRIBE=$VERGEN_GIT_DESCRIBE
 ENV VERGEN_GIT_DIRTY=$VERGEN_GIT_DIRTY
 
-# Build dependencies
-RUN --mount=type=secret,id=DEPOT_TOKEN,env=SCCACHE_WEBDAV_TOKEN \
-    --mount=type=cache,target=/usr/local/cargo/registry,sharing=shared \
-    --mount=type=cache,target=/usr/local/cargo/git,sharing=shared \
-    --mount=type=cache,target=$SCCACHE_DIR,sharing=shared \
-    cargo chef cook --profile $BUILD_PROFILE --features "$FEATURES" --locked --recipe-path recipe.json --manifest-path $MANIFEST_PATH/Cargo.toml
-
 # Build application
 COPY --exclude=.git . .
 RUN --mount=type=secret,id=DEPOT_TOKEN,env=SCCACHE_WEBDAV_TOKEN \

bin/reth-bench-compare/Cargo.toml

@@ -56,7 +56,7 @@ ctrlc.workspace = true
 shlex.workspace = true
 
 [target.'cfg(unix)'.dependencies]
-nix = { version = "0.29", features = ["signal", "process"] }
+nix = { version = "0.31", features = ["signal", "process"] }
 
 [features]
 default = ["jemalloc"]

bin/reth-bench-compare/src/cli.rs

@@ -274,10 +274,10 @@ impl Args {
     /// Get the default RPC URL for a given chain
     const fn get_default_rpc_url(chain: &Chain) -> &'static str {
         match chain.id() {
-            8453 => "https://base-mainnet.rpc.ithaca.xyz",  // base
+            8453 => "https://base.reth.rs/rpc",             // base
             84532 => "https://base-sepolia.rpc.ithaca.xyz", // base-sepolia
             27082 => "https://rpc.hoodi.ethpandaops.io",    // hoodi
-            _ => "https://reth-ethereum.ithaca.xyz/rpc",    // mainnet and fallback
+            _ => "https://ethereum.reth.rs/rpc",            // mainnet and fallback
         }
     }
 

bin/reth-bench/Cargo.toml

@@ -45,7 +45,7 @@ op-alloy-consensus = { workspace = true, features = ["alloy-compat"] }
 op-alloy-rpc-types-engine = { workspace = true, features = ["serde"] }
 
 # reqwest
-reqwest = { workspace = true, default-features = false, features = ["rustls-tls-native-roots"] }
+reqwest.workspace = true
 
 # tower
 tower.workspace = true

bin/reth-bench/src/bench/generate_big_block.rs

@@ -132,13 +132,24 @@ impl<S: TransactionSource> TransactionCollector<S> {
     /// Collect transactions starting from the given block number.
     ///
     /// Skips blob transactions (type 3) and collects until target gas is reached.
-    /// Returns the collected raw transaction bytes, total gas used, and the next block number.
-    pub async fn collect(&self, start_block: u64) -> eyre::Result<(Vec<Bytes>, u64, u64)> {
-        let mut transactions: Vec<Bytes> = Vec::new();
+    /// Returns a `CollectionResult` with transactions, gas info, and next block.
+    pub async fn collect(&self, start_block: u64) -> eyre::Result<CollectionResult> {
+        self.collect_gas(start_block, self.target_gas).await
+    }
+
+    /// Collect transactions up to a specific gas target.
+    ///
+    /// This is used both for initial collection and for retry top-ups.
+    pub async fn collect_gas(
+        &self,
+        start_block: u64,
+        gas_target: u64,
+    ) -> eyre::Result<CollectionResult> {
+        let mut transactions: Vec<RawTransaction> = Vec::new();
         let mut total_gas: u64 = 0;
         let mut current_block = start_block;
 
-        while total_gas < self.target_gas {
+        while total_gas < gas_target {
             let Some((block_txs, _)) = self.source.fetch_block_transactions(current_block).await?
             else {
                 warn!(block = current_block, "Block not found, stopping");
@@ -151,12 +162,12 @@ impl<S: TransactionSource> TransactionCollector<S> {
                     continue;
                 }
 
-                if total_gas + tx.gas_used <= self.target_gas {
-                    transactions.push(tx.raw);
+                if total_gas + tx.gas_used <= gas_target {
                     total_gas += tx.gas_used;
+                    transactions.push(tx);
                 }
 
-                if total_gas >= self.target_gas {
+                if total_gas >= gas_target {
                     break;
                 }
             }
@@ -164,7 +175,7 @@ impl<S: TransactionSource> TransactionCollector<S> {
             current_block += 1;
 
             // Stop early if remaining gas is under 1M (close enough to target)
-            let remaining_gas = self.target_gas.saturating_sub(total_gas);
+            let remaining_gas = gas_target.saturating_sub(total_gas);
             if remaining_gas < 1_000_000 {
                 break;
             }
@@ -172,12 +183,12 @@ impl<S: TransactionSource> TransactionCollector<S> {
 
         info!(
             total_txs = transactions.len(),
-            total_gas,
+            gas_sent = total_gas,
             next_block = current_block,
             "Finished collecting transactions"
         );
 
-        Ok((transactions, total_gas, current_block))
+        Ok(CollectionResult { transactions, gas_sent: total_gas, next_block: current_block })
     }
 }
 
@@ -252,6 +263,80 @@ struct BuiltPayload {
     envelope: ExecutionPayloadEnvelopeV4,
     block_hash: B256,
     timestamp: u64,
+    /// The actual gas used in the built block.
+    gas_used: u64,
+}
+
+/// Result of collecting transactions from blocks.
+#[derive(Debug)]
+pub struct CollectionResult {
+    /// Collected transactions with their gas info.
+    pub transactions: Vec<RawTransaction>,
+    /// Total gas sent (sum of historical `gas_used` for all collected txs).
+    pub gas_sent: u64,
+    /// Next block number to continue collecting from.
+    pub next_block: u64,
+}
+
+/// Constants for retry logic.
+const MAX_BUILD_RETRIES: u32 = 5;
+/// Maximum retries for fetching a transaction batch.
+const MAX_FETCH_RETRIES: u32 = 5;
+/// Tolerance: if `gas_used` is within 1M of target, don't retry.
+const MIN_TARGET_SLACK: u64 = 1_000_000;
+/// Maximum gas to request in retries (10x target as safety cap).
+const MAX_ADDITIONAL_GAS_MULTIPLIER: u64 = 10;
+
+/// Fetches a batch of transactions with retry logic.
+///
+/// Returns `None` if all retries are exhausted.
+async fn fetch_batch_with_retry<S: TransactionSource>(
+    collector: &TransactionCollector<S>,
+    block: u64,
+) -> Option<CollectionResult> {
+    for attempt in 1..=MAX_FETCH_RETRIES {
+        match collector.collect(block).await {
+            Ok(result) => return Some(result),
+            Err(e) => {
+                if attempt == MAX_FETCH_RETRIES {
+                    warn!(attempt, error = %e, "Failed to fetch transactions after max retries");
+                    return None;
+                }
+                warn!(attempt, error = %e, "Failed to fetch transactions, retrying...");
+                tokio::time::sleep(std::time::Duration::from_secs(1)).await;
+            }
+        }
+    }
+    None
+}
+
+/// Outcome of a build attempt check.
+enum RetryOutcome {
+    /// Payload is close enough to target gas.
+    Success,
+    /// Max retries reached, accept what we have.
+    MaxRetries,
+    /// Need more transactions with the specified gas amount.
+    NeedMore(u64),
+}
+
+/// Buffer for receiving transaction batches from the fetcher.
+///
+/// This abstracts over the channel to allow the main loop to request
+/// batches on demand, including for retries.
+struct TxBuffer {
+    receiver: mpsc::Receiver<CollectionResult>,
+}
+
+impl TxBuffer {
+    const fn new(receiver: mpsc::Receiver<CollectionResult>) -> Self {
+        Self { receiver }
+    }
+
+    /// Take the next available batch from the fetcher.
+    async fn take_batch(&mut self) -> Option<CollectionResult> {
+        self.receiver.recv().await
+    }
 }
 
 impl Command {
@@ -312,19 +397,20 @@ impl Command {
             )
             .await?;
         } else {
-            // Single payload - collect transactions and build
+            // Single payload - collect transactions and build with retry
             let tx_source = RpcTransactionSource::from_url(&self.rpc_url)?;
             let collector = TransactionCollector::new(tx_source, self.target_gas);
-            let (transactions, _total_gas, _next_block) = collector.collect(start_block).await?;
+            let result = collector.collect(start_block).await?;
 
-            if transactions.is_empty() {
+            if result.transactions.is_empty() {
                 return Err(eyre::eyre!("No transactions collected"));
             }
 
-            self.execute_sequential(
+            self.execute_sequential_with_retry(
                 &auth_provider,
                 &testing_provider,
-                transactions,
+                &collector,
+                result,
                 parent_hash,
                 parent_timestamp,
             )
@@ -335,32 +421,34 @@ impl Command {
         Ok(())
     }
 
-    /// Sequential execution path for single payload or no-execute mode.
-    async fn execute_sequential(
+    /// Sequential execution path with retry logic for underfilled payloads.
+    async fn execute_sequential_with_retry<S: TransactionSource>(
         &self,
         auth_provider: &RootProvider<AnyNetwork>,
         testing_provider: &RootProvider<AnyNetwork>,
-        transactions: Vec<Bytes>,
+        collector: &TransactionCollector<S>,
+        initial_result: CollectionResult,
         mut parent_hash: B256,
         mut parent_timestamp: u64,
     ) -> eyre::Result<()> {
-        for i in 0..self.count {
-            info!(
-                payload = i + 1,
-                total = self.count,
-                parent_hash = %parent_hash,
-                parent_timestamp = parent_timestamp,
-                "Building payload via testing_buildBlockV1"
-            );
+        let mut current_result = initial_result;
 
+        for i in 0..self.count {
             let built = self
-                .build_payload(testing_provider, &transactions, i, parent_hash, parent_timestamp)
+                .build_with_retry(
+                    testing_provider,
+                    collector,
+                    &mut current_result,
+                    i,
+                    parent_hash,
+                    parent_timestamp,
+                )
                 .await?;
 
             self.save_payload(&built)?;
 
             if self.execute || self.count > 1 {
-                info!(payload = i + 1, block_hash = %built.block_hash, "Executing payload (newPayload + FCU)");
+                info!(payload = i + 1, block_hash = %built.block_hash, gas_used = built.gas_used, "Executing payload (newPayload + FCU)");
                 self.execute_payload_v4(auth_provider, built.envelope, parent_hash).await?;
                 info!(payload = i + 1, "Payload executed successfully");
             }
@@ -371,7 +459,62 @@ impl Command {
         Ok(())
     }
 
-    /// Pipelined execution - fetches transactions and builds payloads in background.
+    /// Build a payload with retry logic when `gas_used` is below target.
+    ///
+    /// Uses the ratio of `gas_used/gas_sent` to estimate how many more transactions
+    /// are needed to hit the target gas.
+    async fn build_with_retry<S: TransactionSource>(
+        &self,
+        testing_provider: &RootProvider<AnyNetwork>,
+        collector: &TransactionCollector<S>,
+        result: &mut CollectionResult,
+        index: u64,
+        parent_hash: B256,
+        parent_timestamp: u64,
+    ) -> eyre::Result<BuiltPayload> {
+        for attempt in 1..=MAX_BUILD_RETRIES {
+            let tx_bytes: Vec<Bytes> = result.transactions.iter().map(|t| t.raw.clone()).collect();
+            let gas_sent = result.gas_sent;
+
+            info!(
+                payload = index + 1,
+                attempt,
+                tx_count = tx_bytes.len(),
+                gas_sent,
+                parent_hash = %parent_hash,
+                "Building payload via testing_buildBlockV1"
+            );
+
+            let built = Self::build_payload_static(
+                testing_provider,
+                &tx_bytes,
+                index,
+                parent_hash,
+                parent_timestamp,
+            )
+            .await?;
+
+            match self.check_retry_outcome(&built, index, attempt, gas_sent) {
+                RetryOutcome::Success | RetryOutcome::MaxRetries => return Ok(built),
+                RetryOutcome::NeedMore(additional_gas) => {
+                    let additional =
+                        collector.collect_gas(result.next_block, additional_gas).await?;
+                    result.transactions.extend(additional.transactions);
+                    result.gas_sent = result.gas_sent.saturating_add(additional.gas_sent);
+                    result.next_block = additional.next_block;
+                }
+            }
+        }
+
+        warn!(payload = index + 1, "Retry loop exited without returning a payload");
+        Err(eyre::eyre!("build_with_retry exhausted retries without result"))
+    }
+
+    /// Pipelined execution - fetches transactions in background, builds with retry.
+    ///
+    /// The fetcher continuously produces transaction batches. The main loop consumes them,
+    /// builds payloads with retry logic (requesting more transactions if underfilled),
+    /// and executes each payload before moving to the next.
     async fn execute_pipelined(
         &self,
         auth_provider: &RootProvider<AnyNetwork>,
@@ -380,167 +523,229 @@ impl Command {
         initial_parent_hash: B256,
         initial_parent_timestamp: u64,
     ) -> eyre::Result<()> {
-        // Create channel for transaction batches (one batch per payload)
-        let (tx_sender, mut tx_receiver) = mpsc::channel::<Vec<Bytes>>(self.prefetch_buffer);
+        // Create channel for transaction batches - fetcher sends CollectionResult
+        let (tx_sender, tx_receiver) = mpsc::channel::<CollectionResult>(self.prefetch_buffer);
 
         // Spawn background task to continuously fetch transaction batches
         let rpc_url = self.rpc_url.clone();
         let target_gas = self.target_gas;
-        let count = self.count;
 
         let fetcher_handle = tokio::spawn(async move {
             let tx_source = match RpcTransactionSource::from_url(&rpc_url) {
                 Ok(source) => source,
                 Err(e) => {
                     warn!(error = %e, "Failed to create transaction source");
-                    return;
+                    return None;
                 }
             };
 
             let collector = TransactionCollector::new(tx_source, target_gas);
             let mut current_block = start_block;
 
-            for payload_idx in 0..count {
-                match collector.collect(current_block).await {
-                    Ok((transactions, total_gas, next_block)) => {
-                        info!(
-                            payload = payload_idx + 1,
-                            tx_count = transactions.len(),
-                            total_gas,
-                            blocks = format!("{}..{}", current_block, next_block),
-                            "Fetched transactions"
-                        );
-                        current_block = next_block;
-
-                        if tx_sender.send(transactions).await.is_err() {
-                            break;
-                        }
-                    }
-                    Err(e) => {
-                        warn!(payload = payload_idx + 1, error = %e, "Failed to fetch transactions");
-                        break;
-                    }
-                }
-            }
-        });
-
-        let mut parent_hash = initial_parent_hash;
-        let mut parent_timestamp = initial_parent_timestamp;
-        let mut pending_build: Option<tokio::task::JoinHandle<eyre::Result<BuiltPayload>>> = None;
-
-        for i in 0..self.count {
-            let is_last = i == self.count - 1;
-
-            // Get current payload (either from pending build or build now)
-            let current_payload = if let Some(handle) = pending_build.take() {
-                handle.await??
-            } else {
-                // First payload - wait for transactions and build synchronously
-                let transactions = tx_receiver
-                    .recv()
-                    .await
-                    .ok_or_else(|| eyre::eyre!("Transaction fetcher stopped unexpectedly"))?;
-
-                if transactions.is_empty() {
-                    return Err(eyre::eyre!("No transactions collected for payload {}", i + 1));
+            while let Some(batch) = fetch_batch_with_retry(&collector, current_block).await {
+                if batch.transactions.is_empty() {
+                    info!(block = current_block, "Reached chain tip, stopping fetcher");
+                    break;
                 }
 
                 info!(
-                    payload = i + 1,
-                    total = self.count,
-                    parent_hash = %parent_hash,
-                    parent_timestamp = parent_timestamp,
-                    tx_count = transactions.len(),
-                    "Building payload via testing_buildBlockV1"
+                    tx_count = batch.transactions.len(),
+                    gas_sent = batch.gas_sent,
+                    blocks = format!("{}..{}", current_block, batch.next_block),
+                    "Fetched transaction batch"
                 );
-                self.build_payload(
+                current_block = batch.next_block;
+
+                if tx_sender.send(batch).await.is_err() {
+                    break;
+                }
+            }
+
+            Some(current_block)
+        });
+
+        // Transaction buffer: holds transactions from batches + any extras from retries
+        let mut tx_buffer = TxBuffer::new(tx_receiver);
+
+        let mut parent_hash = initial_parent_hash;
+        let mut parent_timestamp = initial_parent_timestamp;
+
+        for i in 0..self.count {
+            // Get initial batch of transactions for this payload
+            let Some(mut result) = tx_buffer.take_batch().await else {
+                info!(
+                    payloads_built = i,
+                    payloads_requested = self.count,
+                    "Transaction source exhausted, stopping"
+                );
+                break;
+            };
+
+            if result.transactions.is_empty() {
+                info!(
+                    payloads_built = i,
+                    payloads_requested = self.count,
+                    "No more transactions available, stopping"
+                );
+                break;
+            }
+
+            // Build with retry - may need to request more transactions
+            let built = self
+                .build_with_retry_buffered(
                     testing_provider,
-                    &transactions,
+                    &mut tx_buffer,
+                    &mut result,
                     i,
                     parent_hash,
                     parent_timestamp,
                 )
-                .await?
-            };
+                .await?;
 
-            self.save_payload(&current_payload)?;
+            self.save_payload(&built)?;
 
-            let current_block_hash = current_payload.block_hash;
-            let current_timestamp = current_payload.timestamp;
+            let current_block_hash = built.block_hash;
+            let current_timestamp = built.timestamp;
 
-            // Execute current payload first
-            info!(payload = i + 1, block_hash = %current_block_hash, "Executing payload (newPayload + FCU)");
-            self.execute_payload_v4(auth_provider, current_payload.envelope, parent_hash).await?;
+            // Execute payload
+            info!(payload = i + 1, block_hash = %current_block_hash, gas_used = built.gas_used, "Executing payload (newPayload + FCU)");
+            self.execute_payload_v4(auth_provider, built.envelope, parent_hash).await?;
             info!(payload = i + 1, "Payload executed successfully");
 
-            // Start building next payload in background (if not last) - AFTER execution
-            if !is_last {
-                // Get transactions for next payload (should already be fetched or fetching)
-                let next_transactions = tx_receiver
-                    .recv()
-                    .await
-                    .ok_or_else(|| eyre::eyre!("Transaction fetcher stopped unexpectedly"))?;
-
-                if next_transactions.is_empty() {
-                    return Err(eyre::eyre!("No transactions collected for payload {}", i + 2));
-                }
-
-                let testing_provider = testing_provider.clone();
-                let next_index = i + 1;
-                let total = self.count;
-
-                pending_build = Some(tokio::spawn(async move {
-                    info!(
-                        payload = next_index + 1,
-                        total = total,
-                        parent_hash = %current_block_hash,
-                        parent_timestamp = current_timestamp,
-                        tx_count = next_transactions.len(),
-                        "Building payload via testing_buildBlockV1"
-                    );
-
-                    Self::build_payload_static(
-                        &testing_provider,
-                        &next_transactions,
-                        next_index,
-                        current_block_hash,
-                        current_timestamp,
-                    )
-                    .await
-                }));
-            }
-
             parent_hash = current_block_hash;
             parent_timestamp = current_timestamp;
         }
 
         // Clean up the fetcher task
-        drop(tx_receiver);
+        drop(tx_buffer);
         let _ = fetcher_handle.await;
 
         Ok(())
     }
 
-    /// Build a single payload via `testing_buildBlockV1`.
-    async fn build_payload(
+    /// Build a payload with retry logic, using the buffered transaction source.
+    async fn build_with_retry_buffered(
         &self,
         testing_provider: &RootProvider<AnyNetwork>,
-        transactions: &[Bytes],
+        tx_buffer: &mut TxBuffer,
+        result: &mut CollectionResult,
         index: u64,
         parent_hash: B256,
         parent_timestamp: u64,
     ) -> eyre::Result<BuiltPayload> {
-        Self::build_payload_static(
-            testing_provider,
-            transactions,
-            index,
-            parent_hash,
-            parent_timestamp,
-        )
-        .await
+        for attempt in 1..=MAX_BUILD_RETRIES {
+            let tx_bytes: Vec<Bytes> = result.transactions.iter().map(|t| t.raw.clone()).collect();
+            let gas_sent = result.gas_sent;
+
+            info!(
+                payload = index + 1,
+                attempt,
+                tx_count = tx_bytes.len(),
+                gas_sent,
+                parent_hash = %parent_hash,
+                "Building payload via testing_buildBlockV1"
+            );
+
+            let built = Self::build_payload_static(
+                testing_provider,
+                &tx_bytes,
+                index,
+                parent_hash,
+                parent_timestamp,
+            )
+            .await?;
+
+            match self.check_retry_outcome(&built, index, attempt, gas_sent) {
+                RetryOutcome::Success | RetryOutcome::MaxRetries => return Ok(built),
+                RetryOutcome::NeedMore(additional_gas) => {
+                    let mut collected_gas = 0u64;
+                    while collected_gas < additional_gas {
+                        if let Some(batch) = tx_buffer.take_batch().await {
+                            collected_gas += batch.gas_sent;
+                            result.transactions.extend(batch.transactions);
+                            result.gas_sent = result.gas_sent.saturating_add(batch.gas_sent);
+                            result.next_block = batch.next_block;
+                        } else {
+                            warn!("Transaction fetcher exhausted, proceeding with available transactions");
+                            break;
+                        }
+                    }
+                }
+            }
+        }
+
+        warn!(payload = index + 1, "Retry loop exited without returning a payload");
+        Err(eyre::eyre!("build_with_retry_buffered exhausted retries without result"))
     }
 
-    /// Static version for use in spawned tasks.
+    /// Determines the outcome of a build attempt.
+    fn check_retry_outcome(
+        &self,
+        built: &BuiltPayload,
+        index: u64,
+        attempt: u32,
+        gas_sent: u64,
+    ) -> RetryOutcome {
+        let gas_used = built.gas_used;
+
+        if gas_used + MIN_TARGET_SLACK >= self.target_gas {
+            info!(
+                payload = index + 1,
+                gas_used,
+                target_gas = self.target_gas,
+                attempts = attempt,
+                "Payload built successfully"
+            );
+            return RetryOutcome::Success;
+        }
+
+        if attempt == MAX_BUILD_RETRIES {
+            warn!(
+                payload = index + 1,
+                gas_used,
+                target_gas = self.target_gas,
+                gas_sent,
+                "Underfilled after max retries, accepting payload"
+            );
+            return RetryOutcome::MaxRetries;
+        }
+
+        if gas_used == 0 {
+            warn!(
+                payload = index + 1,
+                "Zero gas used in payload, requesting fixed chunk of additional transactions"
+            );
+            return RetryOutcome::NeedMore(self.target_gas);
+        }
+
+        let gas_sent_needed_total =
+            (self.target_gas as u128 * gas_sent as u128).div_ceil(gas_used as u128) as u64;
+        let additional = gas_sent_needed_total.saturating_sub(gas_sent);
+        let additional = additional.min(self.target_gas * MAX_ADDITIONAL_GAS_MULTIPLIER);
+
+        if additional == 0 {
+            info!(
+                payload = index + 1,
+                gas_used,
+                target_gas = self.target_gas,
+                "No additional transactions needed based on ratio"
+            );
+            return RetryOutcome::Success;
+        }
+
+        let ratio = gas_used as f64 / gas_sent as f64;
+        info!(
+            payload = index + 1,
+            gas_used,
+            gas_sent,
+            ratio = format!("{:.4}", ratio),
+            additional_gas = additional,
+            "Underfilled, collecting more transactions for retry"
+        );
+        RetryOutcome::NeedMore(additional)
+    }
+
+    /// Build a single payload via `testing_buildBlockV1`.
     async fn build_payload_static(
         testing_provider: &RootProvider<AnyNetwork>,
         transactions: &[Bytes],
@@ -578,8 +783,9 @@ impl Command {
         let block_hash = inner.block_hash;
         let block_number = inner.block_number;
         let timestamp = inner.timestamp;
+        let gas_used = inner.gas_used;
 
-        Ok(BuiltPayload { block_number, envelope: v4_envelope, block_hash, timestamp })
+        Ok(BuiltPayload { block_number, envelope: v4_envelope, block_hash, timestamp, gas_used })
     }
 
     /// Save a payload to disk.

bin/reth-bench/src/bench/helpers.rs

@@ -1,6 +1,33 @@
 //! Common helpers for reth-bench commands.
 
 use crate::valid_payload::call_forkchoice_updated;
+use eyre::Result;
+use std::io::{BufReader, Read};
+
+/// Read input from either a file path or stdin.
+pub(crate) fn read_input(path: Option<&str>) -> Result<String> {
+    Ok(match path {
+        Some(path) => reth_fs_util::read_to_string(path)?,
+        None => String::from_utf8(
+            BufReader::new(std::io::stdin()).bytes().collect::<Result<Vec<_>, _>>()?,
+        )?,
+    })
+}
+
+/// Load JWT secret from either a file or use the provided string directly.
+pub(crate) fn load_jwt_secret(jwt_secret: Option<&str>) -> Result<Option<String>> {
+    match jwt_secret {
+        Some(secret) => {
+            // Try to read as file first
+            match std::fs::read_to_string(secret) {
+                Ok(contents) => Ok(Some(contents.trim().to_string())),
+                // If file read fails, use the string directly
+                Err(_) => Ok(Some(secret.to_string())),
+            }
+        }
+        None => Ok(None),
+    }
+}
 
 /// Parses a gas limit value with optional suffix: K for thousand, M for million, G for billion.
 ///

bin/reth-bench/src/bench/mod.rs

@@ -15,6 +15,7 @@ pub use generate_big_block::{
 mod new_payload_fcu;
 mod new_payload_only;
 mod output;
+mod persistence_waiter;
 mod replay_payloads;
 mod send_invalid_payload;
 mod send_payload;

bin/reth-bench/src/bench/new_payload_fcu.rs

@@ -15,28 +15,23 @@ use crate::{
         output::{
             write_benchmark_results, CombinedResult, NewPayloadResult, TotalGasOutput, TotalGasRow,
         },
+        persistence_waiter::{
+            derive_ws_rpc_url, setup_persistence_subscription, PersistenceWaiter,
+            PERSISTENCE_CHECKPOINT_TIMEOUT,
+        },
     },
     valid_payload::{block_to_new_payload, call_forkchoice_updated, call_new_payload},
 };
-use alloy_eips::BlockNumHash;
-use alloy_network::Ethereum;
-use alloy_provider::{Provider, RootProvider};
-use alloy_pubsub::SubscriptionStream;
-use alloy_rpc_client::RpcClient;
+use alloy_provider::Provider;
 use alloy_rpc_types_engine::ForkchoiceState;
-use alloy_transport_ws::WsConnect;
 use clap::Parser;
 use eyre::{Context, OptionExt};
-use futures::StreamExt;
 use humantime::parse_duration;
 use reth_cli_runner::CliContext;
 use reth_engine_primitives::config::DEFAULT_PERSISTENCE_THRESHOLD;
 use reth_node_core::args::BenchmarkArgs;
 use std::time::{Duration, Instant};
 use tracing::{debug, info};
-use url::Url;
-
-const PERSISTENCE_CHECKPOINT_TIMEOUT: Duration = Duration::from_secs(60);
 
 /// `reth benchmark new-payload-fcu` command
 #[derive(Debug, Parser)]
@@ -105,7 +100,11 @@ impl Command {
         let mut waiter = match (self.wait_time, self.wait_for_persistence) {
             (Some(duration), _) => Some(PersistenceWaiter::with_duration(duration)),
             (None, true) => {
-                let sub = self.setup_persistence_subscription().await?;
+                let ws_url = derive_ws_rpc_url(
+                    self.benchmark.ws_rpc_url.as_deref(),
+                    &self.benchmark.engine_rpc_url,
+                )?;
+                let sub = setup_persistence_subscription(ws_url).await?;
                 Some(PersistenceWaiter::with_subscription(
                     sub,
                     self.persistence_threshold,
@@ -245,293 +244,22 @@ impl Command {
             results.into_iter().unzip();
 
         if let Some(ref path) = self.benchmark.output {
-            write_benchmark_results(path, &gas_output_results, combined_results)?;
+            write_benchmark_results(path, &gas_output_results, &combined_results)?;
         }
 
-        let gas_output = TotalGasOutput::new(gas_output_results)?;
+        let gas_output =
+            TotalGasOutput::with_combined_results(gas_output_results, &combined_results)?;
 
         info!(
-            total_duration=?gas_output.total_duration,
-            total_gas_used=?gas_output.total_gas_used,
-            blocks_processed=?gas_output.blocks_processed,
-            "Total Ggas/s: {:.4}",
-            gas_output.total_gigagas_per_second()
+            total_gas_used = gas_output.total_gas_used,
+            total_duration = ?gas_output.total_duration,
+            execution_duration = ?gas_output.execution_duration,
+            blocks_processed = gas_output.blocks_processed,
+            wall_clock_ggas_per_second = format_args!("{:.4}", gas_output.total_gigagas_per_second()),
+            execution_ggas_per_second = format_args!("{:.4}", gas_output.execution_gigagas_per_second()),
+            "Benchmark complete"
         );
 
         Ok(())
     }
-
-    /// Returns the websocket RPC URL used for the persistence subscription.
-    ///
-    /// Preference:
-    /// - If `--ws-rpc-url` is provided, use it directly.
-    /// - Otherwise, derive a WS RPC URL from `--engine-rpc-url`.
-    ///
-    /// The persistence subscription endpoint (`reth_subscribePersistedBlock`) is exposed on
-    /// the regular RPC server (WS port, usually 8546), not on the engine API port (usually 8551).
-    /// Since `BenchmarkArgs` only has the engine URL by default, we convert the scheme
-    /// (http→ws, https→wss) and force the port to 8546.
-    fn derive_ws_rpc_url(&self) -> eyre::Result<Url> {
-        if let Some(ref ws_url) = self.benchmark.ws_rpc_url {
-            let parsed: Url = ws_url
-                .parse()
-                .wrap_err_with(|| format!("Failed to parse WebSocket RPC URL: {ws_url}"))?;
-            info!(target: "reth-bench", ws_url = %parsed, "Using provided WebSocket RPC URL");
-            Ok(parsed)
-        } else {
-            let derived = engine_url_to_ws_url(&self.benchmark.engine_rpc_url)?;
-            debug!(
-                target: "reth-bench",
-                engine_url = %self.benchmark.engine_rpc_url,
-                %derived,
-                "Derived WebSocket RPC URL from engine RPC URL"
-            );
-            Ok(derived)
-        }
-    }
-
-    /// Establishes a websocket connection and subscribes to `reth_subscribePersistedBlock`.
-    async fn setup_persistence_subscription(&self) -> eyre::Result<PersistenceSubscription> {
-        let ws_url = self.derive_ws_rpc_url()?;
-
-        info!("Connecting to WebSocket at {} for persistence subscription", ws_url);
-
-        let ws_connect = WsConnect::new(ws_url.to_string());
-        let client = RpcClient::connect_pubsub(ws_connect)
-            .await
-            .wrap_err("Failed to connect to WebSocket RPC endpoint")?;
-        let provider: RootProvider<Ethereum> = RootProvider::new(client);
-
-        let subscription = provider
-            .subscribe_to::<BlockNumHash>("reth_subscribePersistedBlock")
-            .await
-            .wrap_err("Failed to subscribe to persistence notifications")?;
-
-        info!("Subscribed to persistence notifications");
-
-        Ok(PersistenceSubscription::new(provider, subscription.into_stream()))
-    }
-}
-
-/// Converts an engine API URL to the default RPC websocket URL.
-///
-/// Transformations:
-/// - `http`  → `ws`
-/// - `https` → `wss`
-/// - `ws` / `wss` keep their scheme
-/// - Port is always set to `8546`, reth's default RPC websocket port.
-///
-/// This is used when we only know the engine API URL (typically `:8551`) but
-/// need to connect to the node's WS RPC endpoint for persistence events.
-fn engine_url_to_ws_url(engine_url: &str) -> eyre::Result<Url> {
-    let url: Url = engine_url
-        .parse()
-        .wrap_err_with(|| format!("Failed to parse engine RPC URL: {engine_url}"))?;
-
-    let mut ws_url = url.clone();
-
-    match ws_url.scheme() {
-        "http" => ws_url
-            .set_scheme("ws")
-            .map_err(|_| eyre::eyre!("Failed to set WS scheme for URL: {url}"))?,
-        "https" => ws_url
-            .set_scheme("wss")
-            .map_err(|_| eyre::eyre!("Failed to set WSS scheme for URL: {url}"))?,
-        "ws" | "wss" => {}
-        scheme => {
-            return Err(eyre::eyre!(
-            "Unsupported URL scheme '{scheme}' for URL: {url}. Expected http, https, ws, or wss."
-        ))
-        }
-    }
-
-    ws_url.set_port(Some(8546)).map_err(|_| eyre::eyre!("Failed to set port for URL: {url}"))?;
-
-    Ok(ws_url)
-}
-
-/// Waits until the persistence subscription reports that `target` has been persisted.
-///
-/// Consumes subscription events until `last_persisted >= target`, or returns an error if:
-/// - the subscription stream ends unexpectedly, or
-/// - `timeout` elapses before `target` is observed.
-async fn wait_for_persistence(
-    stream: &mut SubscriptionStream<BlockNumHash>,
-    target: u64,
-    last_persisted: &mut u64,
-    timeout: Duration,
-) -> eyre::Result<()> {
-    tokio::time::timeout(timeout, async {
-        while *last_persisted < target {
-            match stream.next().await {
-                Some(persisted) => {
-                    *last_persisted = persisted.number;
-                    debug!(
-                        target: "reth-bench",
-                        persisted_block = ?last_persisted,
-                        "Received persistence notification"
-                    );
-                }
-                None => {
-                    return Err(eyre::eyre!("Persistence subscription closed unexpectedly"));
-                }
-            }
-        }
-        Ok(())
-    })
-    .await
-    .map_err(|_| {
-        eyre::eyre!(
-            "Persistence timeout: target block {} not persisted within {:?}. Last persisted: {}",
-            target,
-            timeout,
-            last_persisted
-        )
-    })?
-}
-
-/// Wrapper that keeps both the subscription stream and the underlying provider alive.
-/// The provider must be kept alive for the subscription to continue receiving events.
-struct PersistenceSubscription {
-    _provider: RootProvider<Ethereum>,
-    stream: SubscriptionStream<BlockNumHash>,
-}
-
-impl PersistenceSubscription {
-    const fn new(
-        provider: RootProvider<Ethereum>,
-        stream: SubscriptionStream<BlockNumHash>,
-    ) -> Self {
-        Self { _provider: provider, stream }
-    }
-
-    const fn stream_mut(&mut self) -> &mut SubscriptionStream<BlockNumHash> {
-        &mut self.stream
-    }
-}
-
-/// Encapsulates the block waiting logic.
-///
-/// Provides a simple `on_block()` interface that handles both:
-/// - Fixed duration waits (when `wait_time` is set)
-/// - Persistence-based waits (when `subscription` is set)
-///
-/// For persistence mode, waits after every `(threshold + 1)` blocks.
-struct PersistenceWaiter {
-    wait_time: Option<Duration>,
-    subscription: Option<PersistenceSubscription>,
-    blocks_sent: u64,
-    last_persisted: u64,
-    threshold: u64,
-    timeout: Duration,
-}
-
-impl PersistenceWaiter {
-    const fn with_duration(wait_time: Duration) -> Self {
-        Self {
-            wait_time: Some(wait_time),
-            subscription: None,
-            blocks_sent: 0,
-            last_persisted: 0,
-            threshold: 0,
-            timeout: Duration::ZERO,
-        }
-    }
-
-    const fn with_subscription(
-        subscription: PersistenceSubscription,
-        threshold: u64,
-        timeout: Duration,
-    ) -> Self {
-        Self {
-            wait_time: None,
-            subscription: Some(subscription),
-            blocks_sent: 0,
-            last_persisted: 0,
-            threshold,
-            timeout,
-        }
-    }
-
-    /// Called once per block. Waits based on the configured mode.
-    #[allow(clippy::manual_is_multiple_of)]
-    async fn on_block(&mut self, block_number: u64) -> eyre::Result<()> {
-        if let Some(wait_time) = self.wait_time {
-            tokio::time::sleep(wait_time).await;
-            return Ok(());
-        }
-
-        let Some(ref mut subscription) = self.subscription else {
-            return Ok(());
-        };
-
-        self.blocks_sent += 1;
-
-        if self.blocks_sent % (self.threshold + 1) == 0 {
-            debug!(
-                target: "reth-bench",
-                target_block = ?block_number,
-                last_persisted = self.last_persisted,
-                blocks_sent = self.blocks_sent,
-                "Waiting for persistence"
-            );
-
-            wait_for_persistence(
-                subscription.stream_mut(),
-                block_number,
-                &mut self.last_persisted,
-                self.timeout,
-            )
-            .await?;
-
-            debug!(
-                target: "reth-bench",
-                persisted = self.last_persisted,
-                "Persistence caught up"
-            );
-        }
-
-        Ok(())
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn test_engine_url_to_ws_url() {
-        // http -> ws, always uses port 8546
-        let result = engine_url_to_ws_url("http://localhost:8551").unwrap();
-        assert_eq!(result.as_str(), "ws://localhost:8546/");
-
-        // https -> wss
-        let result = engine_url_to_ws_url("https://localhost:8551").unwrap();
-        assert_eq!(result.as_str(), "wss://localhost:8546/");
-
-        // Custom engine port still maps to 8546
-        let result = engine_url_to_ws_url("http://localhost:9551").unwrap();
-        assert_eq!(result.port(), Some(8546));
-
-        // Already ws passthrough
-        let result = engine_url_to_ws_url("ws://localhost:8546").unwrap();
-        assert_eq!(result.scheme(), "ws");
-
-        // Invalid inputs
-        assert!(engine_url_to_ws_url("ftp://localhost:8551").is_err());
-        assert!(engine_url_to_ws_url("not a valid url").is_err());
-    }
-
-    #[tokio::test]
-    async fn test_waiter_with_duration() {
-        let mut waiter = PersistenceWaiter::with_duration(Duration::from_millis(1));
-
-        let start = Instant::now();
-        waiter.on_block(1).await.unwrap();
-        waiter.on_block(2).await.unwrap();
-        waiter.on_block(3).await.unwrap();
-
-        // Should have waited ~3ms total
-        assert!(start.elapsed() >= Duration::from_millis(3));
-    }
 }

bin/reth-bench/src/bench/output.rs

@@ -6,7 +6,7 @@ use csv::Writer;
 use eyre::OptionExt;
 use reth_primitives_traits::constants::GIGAGAS;
 use serde::{ser::SerializeStruct, Deserialize, Serialize};
-use std::{path::Path, time::Duration};
+use std::{fs, path::Path, time::Duration};
 use tracing::info;
 
 /// This is the suffix for gas output csv files.
@@ -158,29 +158,58 @@ pub(crate) struct TotalGasRow {
 pub(crate) struct TotalGasOutput {
     /// The total gas used in the benchmark.
     pub(crate) total_gas_used: u64,
-    /// The total duration of the benchmark.
+    /// The total wall-clock duration of the benchmark (includes wait times).
     pub(crate) total_duration: Duration,
-    /// The total gas used per second.
-    pub(crate) total_gas_per_second: f64,
+    /// The total execution-only duration (excludes wait times).
+    pub(crate) execution_duration: Duration,
     /// The number of blocks processed.
     pub(crate) blocks_processed: u64,
 }
 
 impl TotalGasOutput {
-    /// Create a new [`TotalGasOutput`] from a list of [`TotalGasRow`].
+    /// Create a new [`TotalGasOutput`] from gas rows only.
+    ///
+    /// Use this when execution-only timing is not available (e.g., `new_payload_only`).
+    /// `execution_duration` will equal `total_duration`.
     pub(crate) fn new(rows: Vec<TotalGasRow>) -> eyre::Result<Self> {
-        // the duration is obtained from the last row
         let total_duration = rows.last().map(|row| row.time).ok_or_eyre("empty results")?;
         let blocks_processed = rows.len() as u64;
         let total_gas_used: u64 = rows.into_iter().map(|row| row.gas_used).sum();
-        let total_gas_per_second = total_gas_used as f64 / total_duration.as_secs_f64();
 
-        Ok(Self { total_gas_used, total_duration, total_gas_per_second, blocks_processed })
+        Ok(Self {
+            total_gas_used,
+            total_duration,
+            execution_duration: total_duration,
+            blocks_processed,
+        })
     }
 
-    /// Return the total gigagas per second.
+    /// Create a new [`TotalGasOutput`] from gas rows and combined results.
+    ///
+    /// - `rows`: Used for total gas and wall-clock duration
+    /// - `combined_results`: Used for execution-only duration (sum of `total_latency`)
+    pub(crate) fn with_combined_results(
+        rows: Vec<TotalGasRow>,
+        combined_results: &[CombinedResult],
+    ) -> eyre::Result<Self> {
+        let total_duration = rows.last().map(|row| row.time).ok_or_eyre("empty results")?;
+        let blocks_processed = rows.len() as u64;
+        let total_gas_used: u64 = rows.into_iter().map(|row| row.gas_used).sum();
+
+        // Sum execution-only time from combined results
+        let execution_duration: Duration = combined_results.iter().map(|r| r.total_latency).sum();
+
+        Ok(Self { total_gas_used, total_duration, execution_duration, blocks_processed })
+    }
+
+    /// Return the total gigagas per second based on wall-clock time.
     pub(crate) fn total_gigagas_per_second(&self) -> f64 {
-        self.total_gas_per_second / GIGAGAS as f64
+        self.total_gas_used as f64 / self.total_duration.as_secs_f64() / GIGAGAS as f64
+    }
+
+    /// Return the execution-only gigagas per second (excludes wait times).
+    pub(crate) fn execution_gigagas_per_second(&self) -> f64 {
+        self.total_gas_used as f64 / self.execution_duration.as_secs_f64() / GIGAGAS as f64
     }
 }
 
@@ -192,8 +221,10 @@ impl TotalGasOutput {
 pub(crate) fn write_benchmark_results(
     output_dir: &Path,
     gas_results: &[TotalGasRow],
-    combined_results: Vec<CombinedResult>,
+    combined_results: &[CombinedResult],
 ) -> eyre::Result<()> {
+    fs::create_dir_all(output_dir)?;
+
     let output_path = output_dir.join(COMBINED_OUTPUT_SUFFIX);
     info!("Writing engine api call latency output to file: {:?}", output_path);
     let mut writer = Writer::from_path(&output_path)?;

bin/reth-bench/src/bench/persistence_waiter.rs

@@ -0,0 +1,304 @@
+//! Persistence waiting utilities for benchmarks.
+//!
+//! Provides waiting behavior to control benchmark pacing:
+//! - **Fixed duration waits**: Sleep for a fixed time between blocks
+//! - **Persistence-based waits**: Wait for blocks to be persisted using
+//!   `reth_subscribePersistedBlock` subscription
+
+use alloy_eips::BlockNumHash;
+use alloy_network::Ethereum;
+use alloy_provider::{Provider, RootProvider};
+use alloy_pubsub::SubscriptionStream;
+use alloy_rpc_client::RpcClient;
+use alloy_transport_ws::WsConnect;
+use eyre::Context;
+use futures::StreamExt;
+use std::time::Duration;
+use tracing::{debug, info};
+
+/// Default `WebSocket` RPC port for reth.
+const DEFAULT_WS_RPC_PORT: u16 = 8546;
+use url::Url;
+
+/// Default timeout for waiting on persistence.
+pub(crate) const PERSISTENCE_CHECKPOINT_TIMEOUT: Duration = Duration::from_secs(60);
+
+/// Returns the websocket RPC URL used for the persistence subscription.
+///
+/// Preference:
+/// - If `ws_rpc_url` is provided, use it directly.
+/// - Otherwise, derive a WS RPC URL from `engine_rpc_url`.
+///
+/// The persistence subscription endpoint (`reth_subscribePersistedBlock`) is exposed on
+/// the regular RPC server (WS port, usually 8546), not on the engine API port (usually 8551).
+/// Since we may only have the engine URL by default, we convert the scheme
+/// (http→ws, https→wss) and force the port to 8546.
+pub(crate) fn derive_ws_rpc_url(
+    ws_rpc_url: Option<&str>,
+    engine_rpc_url: &str,
+) -> eyre::Result<Url> {
+    if let Some(ws_url) = ws_rpc_url {
+        let parsed: Url = ws_url
+            .parse()
+            .wrap_err_with(|| format!("Failed to parse WebSocket RPC URL: {ws_url}"))?;
+        info!(target: "reth-bench", ws_url = %parsed, "Using provided WebSocket RPC URL");
+        Ok(parsed)
+    } else {
+        let derived = engine_url_to_ws_url(engine_rpc_url)?;
+        debug!(
+            target: "reth-bench",
+            engine_url = %engine_rpc_url,
+            %derived,
+            "Derived WebSocket RPC URL from engine RPC URL"
+        );
+        Ok(derived)
+    }
+}
+
+/// Converts an engine API URL to the default RPC websocket URL.
+///
+/// Transformations:
+/// - `http`  → `ws`
+/// - `https` → `wss`
+/// - `ws` / `wss` keep their scheme
+/// - Port is always set to `8546`, reth's default RPC websocket port.
+///
+/// This is used when we only know the engine API URL (typically `:8551`) but
+/// need to connect to the node's WS RPC endpoint for persistence events.
+fn engine_url_to_ws_url(engine_url: &str) -> eyre::Result<Url> {
+    let url: Url = engine_url
+        .parse()
+        .wrap_err_with(|| format!("Failed to parse engine RPC URL: {engine_url}"))?;
+
+    let mut ws_url = url.clone();
+
+    match ws_url.scheme() {
+        "http" => ws_url
+            .set_scheme("ws")
+            .map_err(|_| eyre::eyre!("Failed to set WS scheme for URL: {url}"))?,
+        "https" => ws_url
+            .set_scheme("wss")
+            .map_err(|_| eyre::eyre!("Failed to set WSS scheme for URL: {url}"))?,
+        "ws" | "wss" => {}
+        scheme => {
+            return Err(eyre::eyre!(
+            "Unsupported URL scheme '{scheme}' for URL: {url}. Expected http, https, ws, or wss."
+        ))
+        }
+    }
+
+    ws_url
+        .set_port(Some(DEFAULT_WS_RPC_PORT))
+        .map_err(|_| eyre::eyre!("Failed to set port for URL: {url}"))?;
+
+    Ok(ws_url)
+}
+
+/// Waits until the persistence subscription reports that `target` has been persisted.
+///
+/// Consumes subscription events until `last_persisted >= target`, or returns an error if:
+/// - the subscription stream ends unexpectedly, or
+/// - `timeout` elapses before `target` is observed.
+async fn wait_for_persistence(
+    stream: &mut SubscriptionStream<BlockNumHash>,
+    target: u64,
+    last_persisted: &mut u64,
+    timeout: Duration,
+) -> eyre::Result<()> {
+    tokio::time::timeout(timeout, async {
+        while *last_persisted < target {
+            match stream.next().await {
+                Some(persisted) => {
+                    *last_persisted = persisted.number;
+                    debug!(
+                        target: "reth-bench",
+                        persisted_block = ?last_persisted,
+                        "Received persistence notification"
+                    );
+                }
+                None => {
+                    return Err(eyre::eyre!("Persistence subscription closed unexpectedly"));
+                }
+            }
+        }
+        Ok(())
+    })
+    .await
+    .map_err(|_| {
+        eyre::eyre!(
+            "Persistence timeout: target block {} not persisted within {:?}. Last persisted: {}",
+            target,
+            timeout,
+            last_persisted
+        )
+    })?
+}
+
+/// Wrapper that keeps both the subscription stream and the underlying provider alive.
+/// The provider must be kept alive for the subscription to continue receiving events.
+pub(crate) struct PersistenceSubscription {
+    _provider: RootProvider<Ethereum>,
+    stream: SubscriptionStream<BlockNumHash>,
+}
+
+impl PersistenceSubscription {
+    const fn new(
+        provider: RootProvider<Ethereum>,
+        stream: SubscriptionStream<BlockNumHash>,
+    ) -> Self {
+        Self { _provider: provider, stream }
+    }
+
+    const fn stream_mut(&mut self) -> &mut SubscriptionStream<BlockNumHash> {
+        &mut self.stream
+    }
+}
+
+/// Establishes a websocket connection and subscribes to `reth_subscribePersistedBlock`.
+pub(crate) async fn setup_persistence_subscription(
+    ws_url: Url,
+) -> eyre::Result<PersistenceSubscription> {
+    info!("Connecting to WebSocket at {} for persistence subscription", ws_url);
+
+    let ws_connect = WsConnect::new(ws_url.to_string());
+    let client = RpcClient::connect_pubsub(ws_connect)
+        .await
+        .wrap_err("Failed to connect to WebSocket RPC endpoint")?;
+    let provider: RootProvider<Ethereum> = RootProvider::new(client);
+
+    let subscription = provider
+        .subscribe_to::<BlockNumHash>("reth_subscribePersistedBlock")
+        .await
+        .wrap_err("Failed to subscribe to persistence notifications")?;
+
+    info!("Subscribed to persistence notifications");
+
+    Ok(PersistenceSubscription::new(provider, subscription.into_stream()))
+}
+
+/// Encapsulates the block waiting logic.
+///
+/// Provides a simple `on_block()` interface that handles both:
+/// - Fixed duration waits (when `wait_time` is set)
+/// - Persistence-based waits (when `subscription` is set)
+///
+/// For persistence mode, waits after every `(threshold + 1)` blocks.
+pub(crate) struct PersistenceWaiter {
+    wait_time: Option<Duration>,
+    subscription: Option<PersistenceSubscription>,
+    blocks_sent: u64,
+    last_persisted: u64,
+    threshold: u64,
+    timeout: Duration,
+}
+
+impl PersistenceWaiter {
+    pub(crate) const fn with_duration(wait_time: Duration) -> Self {
+        Self {
+            wait_time: Some(wait_time),
+            subscription: None,
+            blocks_sent: 0,
+            last_persisted: 0,
+            threshold: 0,
+            timeout: Duration::ZERO,
+        }
+    }
+
+    pub(crate) const fn with_subscription(
+        subscription: PersistenceSubscription,
+        threshold: u64,
+        timeout: Duration,
+    ) -> Self {
+        Self {
+            wait_time: None,
+            subscription: Some(subscription),
+            blocks_sent: 0,
+            last_persisted: 0,
+            threshold,
+            timeout,
+        }
+    }
+
+    /// Called once per block. Waits based on the configured mode.
+    #[allow(clippy::manual_is_multiple_of)]
+    pub(crate) async fn on_block(&mut self, block_number: u64) -> eyre::Result<()> {
+        if let Some(wait_time) = self.wait_time {
+            tokio::time::sleep(wait_time).await;
+            return Ok(());
+        }
+
+        let Some(ref mut subscription) = self.subscription else {
+            return Ok(());
+        };
+
+        self.blocks_sent += 1;
+
+        if self.blocks_sent % (self.threshold + 1) == 0 {
+            debug!(
+                target: "reth-bench",
+                target_block = ?block_number,
+                last_persisted = self.last_persisted,
+                blocks_sent = self.blocks_sent,
+                "Waiting for persistence"
+            );
+
+            wait_for_persistence(
+                subscription.stream_mut(),
+                block_number,
+                &mut self.last_persisted,
+                self.timeout,
+            )
+            .await?;
+
+            debug!(
+                target: "reth-bench",
+                persisted = self.last_persisted,
+                "Persistence caught up"
+            );
+        }
+
+        Ok(())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::time::Instant;
+
+    #[test]
+    fn test_engine_url_to_ws_url() {
+        // http -> ws, always uses port 8546
+        let result = engine_url_to_ws_url("http://localhost:8551").unwrap();
+        assert_eq!(result.as_str(), "ws://localhost:8546/");
+
+        // https -> wss
+        let result = engine_url_to_ws_url("https://localhost:8551").unwrap();
+        assert_eq!(result.as_str(), "wss://localhost:8546/");
+
+        // Custom engine port still maps to 8546
+        let result = engine_url_to_ws_url("http://localhost:9551").unwrap();
+        assert_eq!(result.port(), Some(8546));
+
+        // Already ws passthrough
+        let result = engine_url_to_ws_url("ws://localhost:8546").unwrap();
+        assert_eq!(result.scheme(), "ws");
+
+        // Invalid inputs
+        assert!(engine_url_to_ws_url("ftp://localhost:8551").is_err());
+        assert!(engine_url_to_ws_url("not a valid url").is_err());
+    }
+
+    #[tokio::test]
+    async fn test_waiter_with_duration() {
+        let mut waiter = PersistenceWaiter::with_duration(Duration::from_millis(1));
+
+        let start = Instant::now();
+        waiter.on_block(1).await.unwrap();
+        waiter.on_block(2).await.unwrap();
+        waiter.on_block(3).await.unwrap();
+
+        // Should have waited ~3ms total
+        assert!(start.elapsed() >= Duration::from_millis(3));
+    }
+}

bin/reth-bench/src/bench/replay_payloads.rs

@@ -2,10 +2,27 @@
 //!
 //! This command reads `ExecutionPayloadEnvelopeV4` files from a directory and replays them
 //! in sequence using `newPayload` followed by `forkchoiceUpdated`.
+//!
+//! Supports configurable waiting behavior:
+//! - **`--wait-time`**: Fixed sleep interval between blocks.
+//! - **`--wait-for-persistence`**: Waits for every Nth block to be persisted using the
+//!   `reth_subscribePersistedBlock` subscription, where N matches the engine's persistence
+//!   threshold. This ensures the benchmark doesn't outpace persistence.
+//!
+//! Both options can be used together or independently.
 
 use crate::{
     authenticated_transport::AuthenticatedTransportConnect,
-    bench::output::GasRampPayloadFile,
+    bench::{
+        output::{
+            write_benchmark_results, CombinedResult, GasRampPayloadFile, NewPayloadResult,
+            TotalGasOutput, TotalGasRow,
+        },
+        persistence_waiter::{
+            derive_ws_rpc_url, setup_persistence_subscription, PersistenceWaiter,
+            PERSISTENCE_CHECKPOINT_TIMEOUT,
+        },
+    },
     valid_payload::{call_forkchoice_updated, call_new_payload},
 };
 use alloy_primitives::B256;
@@ -14,11 +31,16 @@ use alloy_rpc_client::ClientBuilder;
 use alloy_rpc_types_engine::{ExecutionPayloadEnvelopeV4, ForkchoiceState, JwtSecret};
 use clap::Parser;
 use eyre::Context;
-use reqwest::Url;
+use humantime::parse_duration;
 use reth_cli_runner::CliContext;
+use reth_engine_primitives::config::DEFAULT_PERSISTENCE_THRESHOLD;
 use reth_node_api::EngineApiMessageVersion;
-use std::path::PathBuf;
+use std::{
+    path::PathBuf,
+    time::{Duration, Instant},
+};
 use tracing::{debug, info};
+use url::Url;
 
 /// `reth bench replay-payloads` command
 ///
@@ -51,6 +73,42 @@ pub struct Command {
     /// These are replayed before the main payloads to warm up the gas limit.
     #[arg(long, value_name = "GAS_RAMP_DIR")]
     gas_ramp_dir: Option<PathBuf>,
+
+    /// Optional output directory for benchmark results (CSV files).
+    #[arg(long, value_name = "OUTPUT")]
+    output: Option<PathBuf>,
+
+    /// How long to wait after a forkchoice update before sending the next payload.
+    #[arg(long, value_name = "WAIT_TIME", value_parser = parse_duration, verbatim_doc_comment)]
+    wait_time: Option<Duration>,
+
+    /// Wait for blocks to be persisted before sending the next batch.
+    ///
+    /// When enabled, waits for every Nth block to be persisted using the
+    /// `reth_subscribePersistedBlock` subscription. This ensures the benchmark
+    /// doesn't outpace persistence.
+    ///
+    /// The subscription uses the regular RPC websocket endpoint (no JWT required).
+    #[arg(long, default_value = "false", verbatim_doc_comment)]
+    wait_for_persistence: bool,
+
+    /// Engine persistence threshold used for deciding when to wait for persistence.
+    ///
+    /// The benchmark waits after every `(threshold + 1)` blocks. By default this
+    /// matches the engine's `DEFAULT_PERSISTENCE_THRESHOLD` (2), so waits occur
+    /// at blocks 3, 6, 9, etc.
+    #[arg(
+        long = "persistence-threshold",
+        value_name = "PERSISTENCE_THRESHOLD",
+        default_value_t = DEFAULT_PERSISTENCE_THRESHOLD,
+        verbatim_doc_comment
+    )]
+    persistence_threshold: u64,
+
+    /// Optional `WebSocket` RPC URL for persistence subscription.
+    /// If not provided, derives from engine RPC URL by changing scheme to ws and port to 8546.
+    #[arg(long, value_name = "WS_RPC_URL", verbatim_doc_comment)]
+    ws_rpc_url: Option<String>,
 }
 
 /// A loaded payload ready for execution.
@@ -78,6 +136,33 @@ impl Command {
     pub async fn execute(self, _ctx: CliContext) -> eyre::Result<()> {
         info!(payload_dir = %self.payload_dir.display(), "Replaying payloads");
 
+        // Log mode configuration
+        if let Some(duration) = self.wait_time {
+            info!("Using wait-time mode with {}ms delay between blocks", duration.as_millis());
+        }
+        if self.wait_for_persistence {
+            info!(
+                "Persistence waiting enabled (waits after every {} blocks to match engine gap > {} behavior)",
+                self.persistence_threshold + 1,
+                self.persistence_threshold
+            );
+        }
+
+        // Set up waiter based on configured options (duration takes precedence)
+        let mut waiter = match (self.wait_time, self.wait_for_persistence) {
+            (Some(duration), _) => Some(PersistenceWaiter::with_duration(duration)),
+            (None, true) => {
+                let ws_url = derive_ws_rpc_url(self.ws_rpc_url.as_deref(), &self.engine_rpc_url)?;
+                let sub = setup_persistence_subscription(ws_url).await?;
+                Some(PersistenceWaiter::with_subscription(
+                    sub,
+                    self.persistence_threshold,
+                    PERSISTENCE_CHECKPOINT_TIMEOUT,
+                ))
+            }
+            (None, false) => None,
+        };
+
         // Set up authenticated engine provider
         let jwt =
             std::fs::read_to_string(&self.jwt_secret).wrap_err("Failed to read JWT secret file")?;
@@ -144,6 +229,11 @@ impl Command {
             call_forkchoice_updated(&auth_provider, payload.version, fcu_state, None).await?;
 
             info!(gas_ramp_payload = i + 1, "Gas ramp payload executed successfully");
+
+            if let Some(w) = &mut waiter {
+                w.on_block(payload.block_number).await?;
+            }
+
             parent_hash = payload.file.block_hash;
         }
 
@@ -151,22 +241,112 @@ impl Command {
             info!(count = gas_ramp_payloads.len(), "All gas ramp payloads replayed");
         }
 
+        let mut results = Vec::new();
+        let total_benchmark_duration = Instant::now();
+
         for (i, payload) in payloads.iter().enumerate() {
-            info!(
+            let envelope = &payload.envelope;
+            let block_hash = payload.block_hash;
+            let execution_payload = &envelope.envelope_inner.execution_payload;
+            let inner_payload = &execution_payload.payload_inner.payload_inner;
+
+            let gas_used = inner_payload.gas_used;
+            let gas_limit = inner_payload.gas_limit;
+            let block_number = inner_payload.block_number;
+            let transaction_count =
+                execution_payload.payload_inner.payload_inner.transactions.len() as u64;
+
+            debug!(
                 payload = i + 1,
                 total = payloads.len(),
                 index = payload.index,
-                block_hash = %payload.block_hash,
+                block_hash = %block_hash,
                 "Executing payload (newPayload + FCU)"
             );
 
-            self.execute_payload_v4(&auth_provider, &payload.envelope, parent_hash).await?;
+            let start = Instant::now();
 
-            info!(payload = i + 1, "Payload executed successfully");
-            parent_hash = payload.block_hash;
+            debug!(
+                method = "engine_newPayloadV4",
+                block_hash = %block_hash,
+                "Sending newPayload"
+            );
+
+            let status = auth_provider
+                .new_payload_v4(
+                    execution_payload.clone(),
+                    vec![],
+                    B256::ZERO,
+                    envelope.execution_requests.to_vec(),
+                )
+                .await?;
+
+            let new_payload_result = NewPayloadResult { gas_used, latency: start.elapsed() };
+
+            if !status.is_valid() {
+                return Err(eyre::eyre!("Payload rejected: {:?}", status));
+            }
+
+            let fcu_state = ForkchoiceState {
+                head_block_hash: block_hash,
+                safe_block_hash: parent_hash,
+                finalized_block_hash: parent_hash,
+            };
+
+            debug!(method = "engine_forkchoiceUpdatedV3", ?fcu_state, "Sending forkchoiceUpdated");
+
+            let fcu_result = auth_provider.fork_choice_updated_v3(fcu_state, None).await?;
+
+            let total_latency = start.elapsed();
+            let fcu_latency = total_latency - new_payload_result.latency;
+
+            let combined_result = CombinedResult {
+                block_number,
+                gas_limit,
+                transaction_count,
+                new_payload_result,
+                fcu_latency,
+                total_latency,
+            };
+
+            let current_duration = total_benchmark_duration.elapsed();
+            info!(%combined_result);
+
+            if let Some(w) = &mut waiter {
+                w.on_block(block_number).await?;
+            }
+
+            let gas_row =
+                TotalGasRow { block_number, transaction_count, gas_used, time: current_duration };
+            results.push((gas_row, combined_result));
+
+            debug!(?status, ?fcu_result, "Payload executed successfully");
+            parent_hash = block_hash;
         }
 
-        info!(count = payloads.len(), "All payloads replayed successfully");
+        // Drop waiter - we don't need to wait for final blocks to persist
+        // since the benchmark goal is measuring Ggas/s of newPayload/FCU, not persistence.
+        drop(waiter);
+
+        let (gas_output_results, combined_results): (Vec<TotalGasRow>, Vec<CombinedResult>) =
+            results.into_iter().unzip();
+
+        if let Some(ref path) = self.output {
+            write_benchmark_results(path, &gas_output_results, &combined_results)?;
+        }
+
+        let gas_output =
+            TotalGasOutput::with_combined_results(gas_output_results, &combined_results)?;
+        info!(
+            total_gas_used = gas_output.total_gas_used,
+            total_duration = ?gas_output.total_duration,
+            execution_duration = ?gas_output.execution_duration,
+            blocks_processed = gas_output.blocks_processed,
+            wall_clock_ggas_per_second = format_args!("{:.4}", gas_output.total_gigagas_per_second()),
+            execution_ggas_per_second = format_args!("{:.4}", gas_output.execution_gigagas_per_second()),
+            "Benchmark complete"
+        );
+
         Ok(())
     }
 
@@ -284,49 +464,4 @@ impl Command {
 
         Ok(payloads)
     }
-
-    async fn execute_payload_v4(
-        &self,
-        provider: &RootProvider<AnyNetwork>,
-        envelope: &ExecutionPayloadEnvelopeV4,
-        parent_hash: B256,
-    ) -> eyre::Result<()> {
-        let block_hash =
-            envelope.envelope_inner.execution_payload.payload_inner.payload_inner.block_hash;
-
-        debug!(
-            method = "engine_newPayloadV4",
-            block_hash = %block_hash,
-            "Sending newPayload"
-        );
-
-        let status = provider
-            .new_payload_v4(
-                envelope.envelope_inner.execution_payload.clone(),
-                vec![],
-                B256::ZERO,
-                envelope.execution_requests.to_vec(),
-            )
-            .await?;
-
-        info!(?status, "newPayloadV4 response");
-
-        if !status.is_valid() {
-            return Err(eyre::eyre!("Payload rejected: {:?}", status));
-        }
-
-        let fcu_state = ForkchoiceState {
-            head_block_hash: block_hash,
-            safe_block_hash: parent_hash,
-            finalized_block_hash: parent_hash,
-        };
-
-        debug!(method = "engine_forkchoiceUpdatedV3", ?fcu_state, "Sending forkchoiceUpdated");
-
-        let fcu_result = provider.fork_choice_updated_v3(fcu_state, None).await?;
-
-        info!(?fcu_result, "forkchoiceUpdatedV3 response");
-
-        Ok(())
-    }
 }

bin/reth-bench/src/bench/send_invalid_payload/mod.rs

@@ -3,6 +3,7 @@
 mod invalidation;
 use invalidation::InvalidationConfig;
 
+use super::helpers::{load_jwt_secret, read_input};
 use alloy_primitives::{Address, B256};
 use alloy_provider::network::AnyRpcBlock;
 use alloy_rpc_types_engine::ExecutionPayload;
@@ -10,7 +11,7 @@ use clap::Parser;
 use eyre::{OptionExt, Result};
 use op_alloy_consensus::OpTxEnvelope;
 use reth_cli_runner::CliContext;
-use std::io::{BufReader, Read, Write};
+use std::io::Write;
 
 /// Command for generating and sending an invalid `engine_newPayload` request.
 ///
@@ -180,27 +181,6 @@ enum Mode {
 }
 
 impl Command {
-    /// Read input from either a file or stdin
-    fn read_input(&self) -> Result<String> {
-        Ok(match &self.path {
-            Some(path) => reth_fs_util::read_to_string(path)?,
-            None => String::from_utf8(
-                BufReader::new(std::io::stdin()).bytes().collect::<Result<Vec<_>, _>>()?,
-            )?,
-        })
-    }
-
-    /// Load JWT secret from either a file or use the provided string directly
-    fn load_jwt_secret(&self) -> Result<Option<String>> {
-        match &self.jwt_secret {
-            Some(secret) => match std::fs::read_to_string(secret) {
-                Ok(contents) => Ok(Some(contents.trim().to_string())),
-                Err(_) => Ok(Some(secret.clone())),
-            },
-            None => Ok(None),
-        }
-    }
-
     /// Build `InvalidationConfig` from command flags
     const fn build_invalidation_config(&self) -> InvalidationConfig {
         InvalidationConfig {
@@ -236,8 +216,8 @@ impl Command {
 
     /// Execute the command
     pub async fn execute(self, _ctx: CliContext) -> Result<()> {
-        let block_json = self.read_input()?;
-        let jwt_secret = self.load_jwt_secret()?;
+        let block_json = read_input(self.path.as_deref())?;
+        let jwt_secret = load_jwt_secret(self.jwt_secret.as_deref())?;
 
         let block = serde_json::from_str::<AnyRpcBlock>(&block_json)?
             .into_inner()

bin/reth-bench/src/bench/send_payload.rs

@@ -1,10 +1,11 @@
+use super::helpers::{load_jwt_secret, read_input};
 use alloy_provider::network::AnyRpcBlock;
 use alloy_rpc_types_engine::ExecutionPayload;
 use clap::Parser;
 use eyre::{OptionExt, Result};
 use op_alloy_consensus::OpTxEnvelope;
 use reth_cli_runner::CliContext;
-use std::io::{BufReader, Read, Write};
+use std::io::Write;
 
 /// Command for generating and sending an `engine_newPayload` request constructed from an RPC
 /// block.
@@ -51,38 +52,13 @@ enum Mode {
 }
 
 impl Command {
-    /// Read input from either a file or stdin
-    fn read_input(&self) -> Result<String> {
-        Ok(match &self.path {
-            Some(path) => reth_fs_util::read_to_string(path)?,
-            None => String::from_utf8(
-                BufReader::new(std::io::stdin()).bytes().collect::<Result<Vec<_>, _>>()?,
-            )?,
-        })
-    }
-
-    /// Load JWT secret from either a file or use the provided string directly
-    fn load_jwt_secret(&self) -> Result<Option<String>> {
-        match &self.jwt_secret {
-            Some(secret) => {
-                // Try to read as file first
-                match std::fs::read_to_string(secret) {
-                    Ok(contents) => Ok(Some(contents.trim().to_string())),
-                    // If file read fails, use the string directly
-                    Err(_) => Ok(Some(secret.clone())),
-                }
-            }
-            None => Ok(None),
-        }
-    }
-
     /// Execute the generate payload command
     pub async fn execute(self, _ctx: CliContext) -> Result<()> {
         // Load block
-        let block_json = self.read_input()?;
+        let block_json = read_input(self.path.as_deref())?;
 
         // Load JWT secret
-        let jwt_secret = self.load_jwt_secret()?;
+        let jwt_secret = load_jwt_secret(self.jwt_secret.as_deref())?;
 
         // Parse the block
         let block = serde_json::from_str::<AnyRpcBlock>(&block_json)?

bin/reth-bench/src/valid_payload.rs

@@ -260,7 +260,9 @@ pub(crate) async fn call_new_payload<N: Network, P: Provider<N>>(
     while !status.is_valid() {
         if status.is_invalid() {
             error!(?status, ?params, "Invalid {method}",);
-            panic!("Invalid {method}: {status:?}");
+            return Err(alloy_json_rpc::RpcError::LocalUsageError(Box::new(std::io::Error::other(
+                format!("Invalid {method}: {status:?}"),
+            ))))
         }
         if status.is_syncing() {
             return Err(alloy_json_rpc::RpcError::UnsupportedFeature(

crates/chainspec/src/spec.rs

@@ -541,7 +541,7 @@ impl<H: BlockHeader> ChainSpec<H> {
                     }
                 }
 
-                bf_params.first().map(|(_, params)| *params).unwrap_or(BaseFeeParams::ethereum())
+                bf_params.first().map(|(_, params)| *params).unwrap_or_else(BaseFeeParams::ethereum)
             }
         }
     }

crates/cli/commands/Cargo.toml

@@ -78,6 +78,7 @@ lz4.workspace = true
 zstd.workspace = true
 serde.workspace = true
 serde_json.workspace = true
+parking_lot.workspace = true
 tar.workspace = true
 tracing.workspace = true
 backon.workspace = true
@@ -132,4 +133,4 @@ arbitrary = [
     "reth-ethereum-primitives/arbitrary",
 ]
 
-edge = ["reth-db-common/edge", "reth-stages/rocksdb", "reth-provider/rocksdb"]
+edge = ["reth-db-common/edge", "reth-stages/rocksdb", "reth-provider/rocksdb", "reth-prune/rocksdb"]

crates/cli/commands/src/common.rs

@@ -121,13 +121,13 @@ impl<C: ChainSpecParser> EnvironmentArgs<C> {
         let genesis_block_number = self.chain.genesis().number.unwrap_or_default();
         let (db, sfp) = match access {
             AccessRights::RW => (
-                Arc::new(init_db(db_path, self.db.database_args())?),
+                init_db(db_path, self.db.database_args())?,
                 StaticFileProviderBuilder::read_write(sf_path)
                     .with_genesis_block_number(genesis_block_number)
                     .build()?,
             ),
             AccessRights::RO | AccessRights::RoInconsistent => {
-                (Arc::new(open_db_read_only(&db_path, self.db.database_args())?), {
+                (open_db_read_only(&db_path, self.db.database_args())?, {
                     let provider = StaticFileProviderBuilder::read_only(sf_path)
                         .with_genesis_block_number(genesis_block_number)
                         .build()?;
@@ -160,16 +160,16 @@ impl<C: ChainSpecParser> EnvironmentArgs<C> {
     fn create_provider_factory<N: CliNodeTypes>(
         &self,
         config: &Config,
-        db: Arc<DatabaseEnv>,
+        db: DatabaseEnv,
         static_file_provider: StaticFileProvider<N::Primitives>,
         rocksdb_provider: RocksDBProvider,
         access: AccessRights,
-    ) -> eyre::Result<ProviderFactory<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>>
+    ) -> eyre::Result<ProviderFactory<NodeTypesWithDBAdapter<N, DatabaseEnv>>>
     where
         C: ChainSpecParser<ChainSpec = N::ChainSpec>,
     {
         let prune_modes = config.prune.segments.clone();
-        let factory = ProviderFactory::<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>::new(
+        let factory = ProviderFactory::<NodeTypesWithDBAdapter<N, DatabaseEnv>>::new(
             db,
             self.chain.clone(),
             static_file_provider,
@@ -200,7 +200,7 @@ impl<C: ChainSpecParser> EnvironmentArgs<C> {
             let (_tip_tx, tip_rx) = watch::channel(B256::ZERO);
 
             // Builds and executes an unwind-only pipeline
-            let mut pipeline = Pipeline::<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>::builder()
+            let mut pipeline = Pipeline::<NodeTypesWithDBAdapter<N, DatabaseEnv>>::builder()
                 .add_stages(DefaultStages::new(
                     factory.clone(),
                     tip_rx,
@@ -229,7 +229,7 @@ pub struct Environment<N: NodeTypes> {
     /// Configuration for reth node
     pub config: Config,
     /// Provider factory.
-    pub provider_factory: ProviderFactory<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>,
+    pub provider_factory: ProviderFactory<NodeTypesWithDBAdapter<N, DatabaseEnv>>,
     /// Datadir path.
     pub data_dir: ChainPath<DataDirPath>,
 }
@@ -261,8 +261,8 @@ impl AccessRights {
 /// Helper alias to satisfy `FullNodeTypes` bound on [`Node`] trait generic.
 type FullTypesAdapter<T> = FullNodeTypesAdapter<
     T,
-    Arc<DatabaseEnv>,
-    BlockchainProvider<NodeTypesWithDBAdapter<T, Arc<DatabaseEnv>>>,
+    DatabaseEnv,
+    BlockchainProvider<NodeTypesWithDBAdapter<T, DatabaseEnv>>,
 >;
 
 /// Helper trait with a common set of requirements for the

crates/cli/commands/src/db/checksum/mod.rs

@@ -17,7 +17,6 @@ use reth_provider::{providers::ProviderNodeTypes, DBProvider, StaticFileProvider
 use reth_static_file_types::StaticFileSegment;
 use std::{
     hash::{BuildHasher, Hasher},
-    sync::Arc,
     time::{Duration, Instant},
 };
 use tracing::{info, warn};
@@ -90,7 +89,7 @@ impl Command {
     /// Execute `db checksum` command
     pub fn execute<N: CliNodeTypes<ChainSpec: EthereumHardforks>>(
         self,
-        tool: &DbTool<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>,
+        tool: &DbTool<NodeTypesWithDBAdapter<N, DatabaseEnv>>,
     ) -> eyre::Result<()> {
         warn!("This command should be run without the node running!");
 
@@ -117,7 +116,7 @@ fn checksum_hasher() -> impl Hasher {
 }
 
 fn checksum_static_file<N: CliNodeTypes<ChainSpec: EthereumHardforks>>(
-    tool: &DbTool<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>,
+    tool: &DbTool<NodeTypesWithDBAdapter<N, DatabaseEnv>>,
     segment: StaticFileSegment,
     start_block: Option<u64>,
     end_block: Option<u64>,

crates/cli/commands/src/db/checksum/rocksdb.rs

@@ -9,7 +9,7 @@ use reth_db_api::table::Table;
 use reth_db_common::DbTool;
 use reth_node_builder::NodeTypesWithDBAdapter;
 use reth_provider::RocksDBProviderFactory;
-use std::{hash::Hasher, sync::Arc, time::Instant};
+use std::{hash::Hasher, time::Instant};
 use tracing::info;
 
 /// RocksDB tables that can be checksummed.
@@ -36,7 +36,7 @@ impl RocksDbTable {
 
 /// Computes a checksum for a RocksDB table.
 pub fn checksum_rocksdb<N: CliNodeTypes<ChainSpec: EthereumHardforks>>(
-    tool: &DbTool<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>,
+    tool: &DbTool<NodeTypesWithDBAdapter<N, DatabaseEnv>>,
     table: RocksDbTable,
     limit: Option<usize>,
 ) -> eyre::Result<()> {

crates/cli/commands/src/db/diff.rs

@@ -16,7 +16,6 @@ use std::{
     hash::Hash,
     io::Write,
     path::{Path, PathBuf},
-    sync::Arc,
 };
 use tracing::{info, warn};
 
@@ -56,7 +55,7 @@ impl Command {
     /// then written to a file in the output directory.
     pub fn execute<T: NodeTypes>(
         self,
-        tool: &DbTool<NodeTypesWithDBAdapter<T, Arc<DatabaseEnv>>>,
+        tool: &DbTool<NodeTypesWithDBAdapter<T, DatabaseEnv>>,
     ) -> eyre::Result<()> {
         warn!("Make sure the node is not running when running `reth db diff`!");
         // open second db

crates/cli/commands/src/db/list.rs

@@ -7,7 +7,7 @@ use reth_db::{transaction::DbTx, DatabaseEnv};
 use reth_db_api::{database::Database, table::Table, RawValue, TableViewer, Tables};
 use reth_db_common::{DbTool, ListFilter};
 use reth_node_builder::{NodeTypes, NodeTypesWithDBAdapter};
-use std::{cell::RefCell, sync::Arc};
+use std::cell::RefCell;
 use tracing::error;
 
 #[derive(Parser, Debug)]
@@ -55,7 +55,7 @@ impl Command {
     /// Execute `db list` command
     pub fn execute<N: NodeTypes<ChainSpec: EthereumHardforks>>(
         self,
-        tool: &DbTool<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>,
+        tool: &DbTool<NodeTypesWithDBAdapter<N, DatabaseEnv>>,
     ) -> eyre::Result<()> {
         self.table.view(&ListTableViewer { tool, args: &self })
     }
@@ -89,7 +89,7 @@ impl Command {
 }
 
 struct ListTableViewer<'a, N: NodeTypes> {
-    tool: &'a DbTool<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>,
+    tool: &'a DbTool<NodeTypesWithDBAdapter<N, DatabaseEnv>>,
     args: &'a Command,
 }
 
@@ -101,8 +101,8 @@ impl<N: NodeTypes> TableViewer<()> for ListTableViewer<'_, N> {
             // We may be using the tui for a long time
             tx.disable_long_read_transaction_safety();
 
-            let table_db = tx.inner.open_db(Some(self.args.table.name())).wrap_err("Could not open db.")?;
-            let stats = tx.inner.db_stat(table_db.dbi()).wrap_err(format!("Could not find table: {}", self.args.table.name()))?;
+            let table_db = tx.inner().open_db(Some(self.args.table.name())).wrap_err("Could not open db.")?;
+                    let stats = tx.inner().db_stat(table_db.dbi()).wrap_err(format!("Could not find table: {}", self.args.table.name()))?;
             let total_entries = stats.entries();
             let final_entry_idx = total_entries.saturating_sub(1);
             if self.args.skip > final_entry_idx {

crates/cli/commands/src/db/mod.rs

@@ -17,6 +17,7 @@ mod get;
 mod list;
 mod repair_trie;
 mod settings;
+mod state;
 mod static_file_header;
 mod stats;
 /// DB List TUI
@@ -65,6 +66,8 @@ pub enum Subcommands {
     Settings(settings::Command),
     /// Gets storage size information for an account
     AccountStorage(account_storage::Command),
+    /// Gets account state and storage at a specific block
+    State(state::Command),
 }
 
 /// Initializes a provider factory with specified access rights, and then execute with the provided
@@ -198,6 +201,11 @@ impl<C: ChainSpecParser<ChainSpec: EthChainSpec + EthereumHardforks>> Command<C>
                     command.execute(&tool)?;
                 });
             }
+            Subcommands::State(command) => {
+                db_exec!(self.env, tool, N, AccessRights::RO, {
+                    command.execute(&tool)?;
+                });
+            }
         }
 
         Ok(())

crates/cli/commands/src/db/state.rs

@@ -0,0 +1,412 @@
+use alloy_primitives::{Address, BlockNumber, B256, U256};
+use clap::Parser;
+use parking_lot::Mutex;
+use reth_db_api::{
+    cursor::{DbCursorRO, DbDupCursorRO},
+    database::Database,
+    tables,
+    transaction::DbTx,
+};
+use reth_db_common::DbTool;
+use reth_node_builder::NodeTypesWithDB;
+use reth_provider::providers::ProviderNodeTypes;
+use reth_storage_api::{BlockNumReader, StateProvider, StorageSettingsCache};
+use std::{
+    collections::BTreeSet,
+    thread,
+    time::{Duration, Instant},
+};
+use tracing::{error, info};
+
+/// Log progress every 5 seconds
+const LOG_INTERVAL: Duration = Duration::from_secs(30);
+
+/// The arguments for the `reth db state` command
+#[derive(Parser, Debug)]
+pub struct Command {
+    /// The account address to get state for
+    address: Address,
+
+    /// Block number to query state at (uses current state if not provided)
+    #[arg(long, short)]
+    block: Option<BlockNumber>,
+
+    /// Maximum number of storage slots to display
+    #[arg(long, short, default_value = "100")]
+    limit: usize,
+
+    /// Output format (table, json, csv)
+    #[arg(long, short, default_value = "table")]
+    format: OutputFormat,
+}
+
+impl Command {
+    /// Execute `db state` command
+    pub fn execute<N: NodeTypesWithDB + ProviderNodeTypes>(
+        self,
+        tool: &DbTool<N>,
+    ) -> eyre::Result<()> {
+        let address = self.address;
+        let limit = self.limit;
+
+        if let Some(block) = self.block {
+            self.execute_historical(tool, address, block, limit)
+        } else {
+            self.execute_current(tool, address, limit)
+        }
+    }
+
+    fn execute_current<N: NodeTypesWithDB + ProviderNodeTypes>(
+        &self,
+        tool: &DbTool<N>,
+        address: Address,
+        limit: usize,
+    ) -> eyre::Result<()> {
+        let entries = tool.provider_factory.db_ref().view(|tx| {
+            // Get account info
+            let account = tx.get::<tables::PlainAccountState>(address)?;
+
+            // Get storage entries
+            let mut cursor = tx.cursor_dup_read::<tables::PlainStorageState>()?;
+            let mut entries = Vec::new();
+            let mut last_log = Instant::now();
+
+            let walker = cursor.walk_dup(Some(address), None)?;
+            for (idx, entry) in walker.enumerate() {
+                let (_, storage_entry) = entry?;
+
+                if storage_entry.value != U256::ZERO {
+                    entries.push((storage_entry.key, storage_entry.value));
+                }
+
+                if entries.len() >= limit {
+                    break;
+                }
+
+                if last_log.elapsed() >= LOG_INTERVAL {
+                    info!(
+                        target: "reth::cli",
+                        address = %address,
+                        slots_scanned = idx,
+                        "Scanning storage slots"
+                    );
+                    last_log = Instant::now();
+                }
+            }
+
+            Ok::<_, eyre::Report>((account, entries))
+        })??;
+
+        let (account, storage_entries) = entries;
+
+        self.print_results(address, None, account, &storage_entries);
+
+        Ok(())
+    }
+
+    fn execute_historical<N: NodeTypesWithDB + ProviderNodeTypes>(
+        &self,
+        tool: &DbTool<N>,
+        address: Address,
+        block: BlockNumber,
+        limit: usize,
+    ) -> eyre::Result<()> {
+        let provider = tool.provider_factory.history_by_block_number(block)?;
+
+        // Get account info at that block
+        let account = provider.basic_account(&address)?;
+
+        // Check storage settings to determine where history is stored
+        let storage_settings = tool.provider_factory.cached_storage_settings();
+        let history_in_rocksdb = storage_settings.storages_history_in_rocksdb;
+
+        // For historical queries, enumerate keys from history indices only
+        // (not PlainStorageState, which reflects current state)
+        let mut storage_keys = BTreeSet::new();
+
+        if history_in_rocksdb {
+            error!(
+                target: "reth::cli",
+                "Historical storage queries with RocksDB backend are not yet supported. \
+                 Use MDBX for storage history or query current state without --block."
+            );
+            return Ok(());
+        }
+
+        // Collect keys from MDBX StorageChangeSets using parallel scanning
+        self.collect_mdbx_storage_keys_parallel(tool, address, &mut storage_keys)?;
+
+        info!(
+            target: "reth::cli",
+            address = %address,
+            block = block,
+            total_keys = storage_keys.len(),
+            "Found storage keys to query"
+        );
+
+        // Now query each key at the historical block using the StateProvider
+        // This handles both MDBX and RocksDB backends transparently
+        let mut entries = Vec::new();
+        let mut last_log = Instant::now();
+
+        for (idx, key) in storage_keys.iter().enumerate() {
+            match provider.storage(address, *key) {
+                Ok(Some(value)) if value != U256::ZERO => {
+                    entries.push((*key, value));
+                }
+                _ => {}
+            }
+
+            if entries.len() >= limit {
+                break;
+            }
+
+            if last_log.elapsed() >= LOG_INTERVAL {
+                info!(
+                    target: "reth::cli",
+                    address = %address,
+                    block = block,
+                    keys_total = storage_keys.len(),
+                    slots_scanned = idx,
+                    slots_found = entries.len(),
+                    "Scanning historical storage slots"
+                );
+                last_log = Instant::now();
+            }
+        }
+
+        self.print_results(address, Some(block), account, &entries);
+
+        Ok(())
+    }
+
+    /// Collects storage keys from MDBX StorageChangeSets using parallel block range scanning.
+    fn collect_mdbx_storage_keys_parallel<N: NodeTypesWithDB + ProviderNodeTypes>(
+        &self,
+        tool: &DbTool<N>,
+        address: Address,
+        keys: &mut BTreeSet<B256>,
+    ) -> eyre::Result<()> {
+        const CHUNK_SIZE: u64 = 500_000; // 500k blocks per thread
+        let num_threads = std::thread::available_parallelism()
+            .map(|p| p.get().saturating_sub(1).max(1))
+            .unwrap_or(4);
+
+        // Get the current tip block
+        let tip = tool.provider_factory.provider()?.best_block_number()?;
+
+        if tip == 0 {
+            return Ok(());
+        }
+
+        info!(
+            target: "reth::cli",
+            address = %address,
+            tip,
+            chunk_size = CHUNK_SIZE,
+            num_threads,
+            "Starting parallel MDBX changeset scan"
+        );
+
+        // Shared state for collecting keys
+        let collected_keys: Mutex<BTreeSet<B256>> = Mutex::new(BTreeSet::new());
+        let total_entries_scanned = Mutex::new(0usize);
+
+        // Create chunk ranges
+        let mut chunks: Vec<(u64, u64)> = Vec::new();
+        let mut start = 0u64;
+        while start <= tip {
+            let end = (start + CHUNK_SIZE - 1).min(tip);
+            chunks.push((start, end));
+            start = end + 1;
+        }
+
+        let chunks_ref = &chunks;
+        let next_chunk = Mutex::new(0usize);
+        let next_chunk_ref = &next_chunk;
+        let collected_keys_ref = &collected_keys;
+        let total_entries_ref = &total_entries_scanned;
+
+        thread::scope(|s| {
+            let handles: Vec<_> = (0..num_threads)
+                .map(|thread_id| {
+                    s.spawn(move || {
+                        loop {
+                            // Get next chunk to process
+                            let chunk_idx = {
+                                let mut idx = next_chunk_ref.lock();
+                                if *idx >= chunks_ref.len() {
+                                    return Ok::<_, eyre::Report>(());
+                                }
+                                let current = *idx;
+                                *idx += 1;
+                                current
+                            };
+
+                            let (chunk_start, chunk_end) = chunks_ref[chunk_idx];
+
+                            // Open a new read transaction for this chunk
+                            tool.provider_factory.db_ref().view(|tx| {
+                                tx.disable_long_read_transaction_safety();
+
+                                let mut changeset_cursor =
+                                    tx.cursor_read::<tables::StorageChangeSets>()?;
+                                let start_key =
+                                    reth_db_api::models::BlockNumberAddress((chunk_start, address));
+                                let end_key =
+                                    reth_db_api::models::BlockNumberAddress((chunk_end, address));
+
+                                let mut local_keys = BTreeSet::new();
+                                let mut entries_in_chunk = 0usize;
+
+                                if let Ok(walker) = changeset_cursor.walk_range(start_key..=end_key)
+                                {
+                                    for (block_addr, storage_entry) in walker.flatten() {
+                                        if block_addr.address() == address {
+                                            local_keys.insert(storage_entry.key);
+                                        }
+                                        entries_in_chunk += 1;
+                                    }
+                                }
+
+                                // Merge into global state
+                                collected_keys_ref.lock().extend(local_keys);
+                                *total_entries_ref.lock() += entries_in_chunk;
+
+                                info!(
+                                    target: "reth::cli",
+                                    thread_id,
+                                    chunk_start,
+                                    chunk_end,
+                                    entries_in_chunk,
+                                    "Thread completed chunk"
+                                );
+
+                                Ok::<_, eyre::Report>(())
+                            })??;
+                        }
+                    })
+                })
+                .collect();
+
+            for handle in handles {
+                handle.join().map_err(|_| eyre::eyre!("Thread panicked"))??;
+            }
+
+            Ok::<_, eyre::Report>(())
+        })?;
+
+        let final_keys = collected_keys.into_inner();
+        let total = *total_entries_scanned.lock();
+
+        info!(
+            target: "reth::cli",
+            address = %address,
+            total_entries = total,
+            unique_keys = final_keys.len(),
+            "Finished parallel MDBX changeset scan"
+        );
+
+        keys.extend(final_keys);
+        Ok(())
+    }
+
+    fn print_results(
+        &self,
+        address: Address,
+        block: Option<BlockNumber>,
+        account: Option<reth_primitives_traits::Account>,
+        storage: &[(alloy_primitives::B256, U256)],
+    ) {
+        match self.format {
+            OutputFormat::Table => {
+                println!("Account: {address}");
+                if let Some(b) = block {
+                    println!("Block: {b}");
+                } else {
+                    println!("Block: latest");
+                }
+                println!();
+
+                if let Some(acc) = account {
+                    println!("Nonce: {}", acc.nonce);
+                    println!("Balance: {} wei", acc.balance);
+                    if let Some(code_hash) = acc.bytecode_hash {
+                        println!("Code hash: {code_hash}");
+                    }
+                } else {
+                    println!("Account not found");
+                }
+
+                println!();
+                println!("Storage ({} slots):", storage.len());
+                println!("{:-<130}", "");
+                println!("{:<66} | {:<64}", "Slot", "Value");
+                println!("{:-<130}", "");
+                for (key, value) in storage {
+                    println!("{key} | {value:#066x}");
+                }
+            }
+            OutputFormat::Json => {
+                let output = serde_json::json!({
+                    "address": address.to_string(),
+                    "block": block,
+                    "account": account.map(|a| serde_json::json!({
+                        "nonce": a.nonce,
+                        "balance": a.balance.to_string(),
+                        "code_hash": a.bytecode_hash.map(|h| h.to_string()),
+                    })),
+                    "storage": storage.iter().map(|(k, v)| {
+                        serde_json::json!({
+                            "key": k.to_string(),
+                            "value": format!("{v:#066x}"),
+                        })
+                    }).collect::<Vec<_>>(),
+                });
+                println!("{}", serde_json::to_string_pretty(&output).unwrap());
+            }
+            OutputFormat::Csv => {
+                println!("slot,value");
+                for (key, value) in storage {
+                    println!("{key},{value:#066x}");
+                }
+            }
+        }
+    }
+}
+
+#[derive(Debug, Clone, Default, clap::ValueEnum)]
+pub enum OutputFormat {
+    #[default]
+    Table,
+    Json,
+    Csv,
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn parse_state_args() {
+        let cmd = Command::try_parse_from([
+            "state",
+            "0xd8dA6BF26964aF9D7eEd9e03E53415D37aA96045",
+            "--block",
+            "1000000",
+        ])
+        .unwrap();
+        assert_eq!(
+            cmd.address,
+            "0xd8dA6BF26964aF9D7eEd9e03E53415D37aA96045".parse::<Address>().unwrap()
+        );
+        assert_eq!(cmd.block, Some(1000000));
+    }
+
+    #[test]
+    fn parse_state_args_no_block() {
+        let cmd = Command::try_parse_from(["state", "0xd8dA6BF26964aF9D7eEd9e03E53415D37aA96045"])
+            .unwrap();
+        assert_eq!(cmd.block, None);
+    }
+}

crates/cli/commands/src/db/stats.rs

@@ -16,7 +16,7 @@ use reth_provider::{
     RocksDBProviderFactory,
 };
 use reth_static_file_types::SegmentRangeInclusive;
-use std::{sync::Arc, time::Duration};
+use std::time::Duration;
 
 #[derive(Parser, Debug)]
 /// The arguments for the `reth db stats` command
@@ -48,7 +48,7 @@ impl Command {
     pub fn execute<N: CliNodeTypes<ChainSpec: EthereumHardforks>>(
         self,
         data_dir: ChainPath<DataDirPath>,
-        tool: &DbTool<NodeTypesWithDBAdapter<N, Arc<DatabaseEnv>>>,
+        tool: &DbTool<NodeTypesWithDBAdapter<N, DatabaseEnv>>,
     ) -> eyre::Result<()> {
         if self.checksum {
             let checksum_report = self.checksum_report(tool)?;
@@ -72,7 +72,7 @@ impl Command {
         Ok(())
     }
 
-    fn db_stats_table<N: NodeTypesWithDB<DB = Arc<DatabaseEnv>>>(
+    fn db_stats_table<N: NodeTypesWithDB<DB = DatabaseEnv>>(
         &self,
         tool: &DbTool<N>,
     ) -> eyre::Result<ComfyTable> {
@@ -92,10 +92,10 @@ impl Command {
             db_tables.sort();
             let mut total_size = 0;
             for db_table in db_tables {
-                let table_db = tx.inner.open_db(Some(db_table)).wrap_err("Could not open db.")?;
+                let table_db = tx.inner().open_db(Some(db_table)).wrap_err("Could not open db.")?;
 
                 let stats = tx
-                    .inner
+                    .inner()
                     .db_stat(table_db.dbi())
                     .wrap_err(format!("Could not find table: {db_table}"))?;
 
@@ -136,9 +136,9 @@ impl Command {
                 .add_cell(Cell::new(human_bytes(total_size as f64)));
             table.add_row(row);
 
-            let freelist = tx.inner.env().freelist()?;
+            let freelist = tx.inner().env().freelist()?;
             let pagesize =
-                tx.inner.db_stat(mdbx::Database::freelist_db().dbi())?.page_size() as usize;
+                tx.inner().db_stat(mdbx::Database::freelist_db().dbi())?.page_size() as usize;
             let freelist_size = freelist * pagesize;
 
             let mut row = Row::new();

crates/cli/commands/src/db/tui.rs

@@ -227,8 +227,9 @@ where
 
         // Handle errors
         if let Err(err) = res {
-            error!("{:?}", err)
+            error!("{err}");
         }
+
         Ok(())
     }
 }
@@ -241,6 +242,7 @@ fn event_loop<B: Backend, F, T: Table>(
 ) -> io::Result<()>
 where
     F: FnMut(usize, usize) -> Vec<TableRow<T>>,
+    io::Error: From<B::Error>,
 {
     let mut last_tick = Instant::now();
     let mut running = true;

crates/cli/commands/src/launcher.rs

@@ -2,7 +2,7 @@ use futures::Future;
 use reth_cli::chainspec::ChainSpecParser;
 use reth_db::DatabaseEnv;
 use reth_node_builder::{NodeBuilder, WithLaunchContext};
-use std::{fmt, sync::Arc};
+use std::fmt;
 
 /// A trait for launching a reth node with custom configuration strategies.
 ///
@@ -30,7 +30,7 @@ where
     /// * `builder_args` - Extension arguments for configuration
     fn entrypoint(
         self,
-        builder: WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+        builder: WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
         builder_args: Ext,
     ) -> impl Future<Output = eyre::Result<()>>;
 }
@@ -58,7 +58,7 @@ impl<F> FnLauncher<F> {
     where
         C: ChainSpecParser,
         F: AsyncFnOnce(
-            WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+            WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
             Ext,
         ) -> eyre::Result<()>,
     {
@@ -77,13 +77,13 @@ where
     C: ChainSpecParser,
     Ext: clap::Args + fmt::Debug,
     F: AsyncFnOnce(
-        WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+        WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
         Ext,
     ) -> eyre::Result<()>,
 {
     fn entrypoint(
         self,
-        builder: WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+        builder: WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
         builder_args: Ext,
     ) -> impl Future<Output = eyre::Result<()>> {
         (self.func)(builder, builder_args)

crates/cli/commands/src/node.rs

@@ -206,7 +206,7 @@ where
         let db_path = data_dir.db();
 
         tracing::info!(target: "reth::cli", path = ?db_path, "Opening database");
-        let database = Arc::new(init_db(db_path.clone(), self.db.database_args())?.with_metrics());
+        let database = init_db(db_path.clone(), self.db.database_args())?.with_metrics();
 
         if with_unused_ports {
             node_config = node_config.with_unused_ports();

crates/cli/commands/src/prune.rs

@@ -1,26 +1,62 @@
-//! Command that runs pruning without any limits.
+//! Command that runs pruning.
 use crate::common::{AccessRights, CliNodeTypes, EnvironmentArgs};
 use clap::Parser;
-use reth_chainspec::{EthChainSpec, EthereumHardforks};
+use reth_chainspec::{ChainSpecProvider, EthChainSpec, EthereumHardforks};
 use reth_cli::chainspec::ChainSpecParser;
+use reth_cli_runner::CliContext;
+use reth_node_builder::common::metrics_hooks;
+use reth_node_core::{args::MetricArgs, version::version_metadata};
+use reth_node_metrics::{
+    chain::ChainSpecInfo,
+    server::{MetricServer, MetricServerConfig},
+    version::VersionInfo,
+};
 use reth_prune::PrunerBuilder;
 use reth_static_file::StaticFileProducer;
 use std::sync::Arc;
 use tracing::info;
 
-/// Prunes according to the configuration without any limits
+/// Prunes according to the configuration
 #[derive(Debug, Parser)]
 pub struct PruneCommand<C: ChainSpecParser> {
     #[command(flatten)]
     env: EnvironmentArgs<C>,
+
+    /// Prometheus metrics configuration.
+    #[command(flatten)]
+    metrics: MetricArgs,
 }
 
 impl<C: ChainSpecParser<ChainSpec: EthChainSpec + EthereumHardforks>> PruneCommand<C> {
     /// Execute the `prune` command
-    pub async fn execute<N: CliNodeTypes<ChainSpec = C::ChainSpec>>(self) -> eyre::Result<()> {
+    pub async fn execute<N: CliNodeTypes<ChainSpec = C::ChainSpec>>(
+        self,
+        ctx: CliContext,
+    ) -> eyre::Result<()> {
         let env = self.env.init::<N>(AccessRights::RW)?;
         let provider_factory = env.provider_factory;
         let config = env.config.prune;
+        let data_dir = env.data_dir;
+
+        if let Some(listen_addr) = self.metrics.prometheus {
+            let config = MetricServerConfig::new(
+                listen_addr,
+                VersionInfo {
+                    version: version_metadata().cargo_pkg_version.as_ref(),
+                    build_timestamp: version_metadata().vergen_build_timestamp.as_ref(),
+                    cargo_features: version_metadata().vergen_cargo_features.as_ref(),
+                    git_sha: version_metadata().vergen_git_sha.as_ref(),
+                    target_triple: version_metadata().vergen_cargo_target_triple.as_ref(),
+                    build_profile: version_metadata().build_profile_name.as_ref(),
+                },
+                ChainSpecInfo { name: provider_factory.chain_spec().chain().to_string() },
+                ctx.task_executor,
+                metrics_hooks(&provider_factory),
+                data_dir.pprof_dumps(),
+            );
+
+            MetricServer::new(config).serve().await?;
+        }
 
         // Copy data from database to static files
         info!(target: "reth::cli", "Copying data from database to static files...");
@@ -33,13 +69,43 @@ impl<C: ChainSpecParser<ChainSpec: EthChainSpec + EthereumHardforks>> PruneComma
         // Delete data which has been copied to static files.
         if let Some(prune_tip) = lowest_static_file_height {
             info!(target: "reth::cli", ?prune_tip, ?config, "Pruning data from database...");
-            // Run the pruner according to the configuration, and don't enforce any limits on it
+
+            // Use batched pruning with a limit to bound memory, running in a loop until complete.
+            const DELETE_LIMIT: usize = 200_000;
             let mut pruner = PrunerBuilder::new(config)
-                .delete_limit(usize::MAX)
+                .delete_limit(DELETE_LIMIT)
                 .build_with_provider_factory(provider_factory);
 
-            pruner.run(prune_tip)?;
-            info!(target: "reth::cli", "Pruned data from database");
+            let mut total_pruned = 0usize;
+            loop {
+                let output = pruner.run(prune_tip)?;
+                let batch_pruned: usize = output.segments.iter().map(|(_, seg)| seg.pruned).sum();
+                total_pruned = total_pruned.saturating_add(batch_pruned);
+
+                // Check if all segments are finished (not just the overall progress,
+                // since the pruner sets overall progress from the last segment only)
+                let all_segments_finished =
+                    output.segments.iter().all(|(_, seg)| seg.progress.is_finished());
+
+                if all_segments_finished {
+                    break;
+                }
+
+                if batch_pruned == 0 {
+                    return Err(eyre::eyre!(
+                        "pruner made no progress but reported more data remaining; \
+                         aborting to prevent infinite loop"
+                    ));
+                }
+
+                info!(
+                    target: "reth::cli",
+                    batch_pruned,
+                    total_pruned,
+                    "Pruning batch complete, continuing..."
+                );
+            }
+            info!(target: "reth::cli", total_pruned, "Pruned data from database");
         }
 
         Ok(())

crates/cli/commands/src/stage/dump/execution.rs

@@ -26,7 +26,7 @@ pub(crate) async fn dump_execution_stage<N, E, C>(
     consensus: C,
 ) -> eyre::Result<()>
 where
-    N: ProviderNodeTypes<DB = Arc<DatabaseEnv>>,
+    N: ProviderNodeTypes<DB = DatabaseEnv>,
     E: ConfigureEvm<Primitives = N::Primitives> + 'static,
     C: FullConsensus<E::Primitives> + 'static,
 {
@@ -39,7 +39,7 @@ where
     if should_run {
         dry_run(
             ProviderFactory::<N>::new(
-                Arc::new(output_db),
+                output_db,
                 db_tool.chain(),
                 StaticFileProvider::read_write(output_datadir.static_files())?,
                 RocksDBProvider::builder(output_datadir.rocksdb()).build()?,

crates/cli/commands/src/stage/dump/hashing_account.rs

@@ -10,10 +10,9 @@ use reth_provider::{
     DatabaseProviderFactory, ProviderFactory,
 };
 use reth_stages::{stages::AccountHashingStage, Stage, StageCheckpoint, UnwindInput};
-use std::sync::Arc;
 use tracing::info;
 
-pub(crate) async fn dump_hashing_account_stage<N: ProviderNodeTypes<DB = Arc<DatabaseEnv>>>(
+pub(crate) async fn dump_hashing_account_stage<N: ProviderNodeTypes<DB = DatabaseEnv>>(
     db_tool: &DbTool<N>,
     from: BlockNumber,
     to: BlockNumber,
@@ -36,7 +35,7 @@ pub(crate) async fn dump_hashing_account_stage<N: ProviderNodeTypes<DB = Arc<Dat
     if should_run {
         dry_run(
             ProviderFactory::<N>::new(
-                Arc::new(output_db),
+                output_db,
                 db_tool.chain(),
                 StaticFileProvider::read_write(output_datadir.static_files())?,
                 RocksDBProvider::builder(output_datadir.rocksdb()).build()?,

crates/cli/commands/src/stage/dump/hashing_storage.rs

@@ -9,10 +9,9 @@ use reth_provider::{
     DatabaseProviderFactory, ProviderFactory,
 };
 use reth_stages::{stages::StorageHashingStage, Stage, StageCheckpoint, UnwindInput};
-use std::sync::Arc;
 use tracing::info;
 
-pub(crate) async fn dump_hashing_storage_stage<N: ProviderNodeTypes<DB = Arc<DatabaseEnv>>>(
+pub(crate) async fn dump_hashing_storage_stage<N: ProviderNodeTypes<DB = DatabaseEnv>>(
     db_tool: &DbTool<N>,
     from: u64,
     to: u64,
@@ -26,7 +25,7 @@ pub(crate) async fn dump_hashing_storage_stage<N: ProviderNodeTypes<DB = Arc<Dat
     if should_run {
         dry_run(
             ProviderFactory::<N>::new(
-                Arc::new(output_db),
+                output_db,
                 db_tool.chain(),
                 StaticFileProvider::read_write(output_datadir.static_files())?,
                 RocksDBProvider::builder(output_datadir.rocksdb()).build()?,

crates/cli/commands/src/stage/dump/merkle.rs

@@ -34,7 +34,7 @@ pub(crate) async fn dump_merkle_stage<N>(
     consensus: impl FullConsensus<N::Primitives> + 'static,
 ) -> Result<()>
 where
-    N: ProviderNodeTypes<DB = Arc<DatabaseEnv>>,
+    N: ProviderNodeTypes<DB = DatabaseEnv>,
 {
     let (output_db, tip_block_number) = setup(from, to, &output_datadir.db(), db_tool)?;
 
@@ -59,7 +59,7 @@ where
     if should_run {
         dry_run(
             ProviderFactory::<N>::new(
-                Arc::new(output_db),
+                output_db,
                 db_tool.chain(),
                 StaticFileProvider::read_write(output_datadir.static_files())?,
                 RocksDBProvider::builder(output_datadir.rocksdb()).build()?,

crates/cli/commands/src/stage/unwind.rs

@@ -158,7 +158,7 @@ enum Subcommands {
 
 impl Subcommands {
     /// Returns the block to unwind to. The returned block will stay in database.
-    fn unwind_target<N: ProviderNodeTypes<DB = Arc<DatabaseEnv>>>(
+    fn unwind_target<N: ProviderNodeTypes<DB = DatabaseEnv>>(
         &self,
         factory: ProviderFactory<N>,
     ) -> eyre::Result<u64> {

crates/consensus/debug-client/Cargo.toml

@@ -29,7 +29,7 @@ auto_impl.workspace = true
 derive_more.workspace = true
 futures.workspace = true
 eyre.workspace = true
-reqwest = { workspace = true, features = ["rustls-tls"] }
+reqwest.workspace = true
 serde = { workspace = true, features = ["derive"] }
 tokio = { workspace = true, features = ["time"] }
 serde_json.workspace = true

crates/consensus/debug-client/src/client.rs

@@ -38,12 +38,8 @@ pub trait BlockProvider: Send + Sync + 'static {
         offset: usize,
     ) -> impl Future<Output = eyre::Result<B256>> + Send {
         async move {
-            let stored_hash = previous_block_hashes
-                .len()
-                .checked_sub(offset)
-                .and_then(|index| previous_block_hashes.get(index));
-            if let Some(hash) = stored_hash {
-                return Ok(*hash);
+            if let Some(hash) = get_hash_at_offset(previous_block_hashes, offset) {
+                return Ok(hash);
             }
 
             // Return zero hash if the chain isn't long enough to have the block at the offset.
@@ -83,7 +79,7 @@ where
     /// Spawn the client to start sending FCUs and new payloads by periodically fetching recent
     /// blocks.
     pub async fn run(self) {
-        let mut previous_block_hashes = AllocRingBuffer::new(64);
+        let mut previous_block_hashes = AllocRingBuffer::new(65);
         let mut block_stream = {
             let (tx, rx) = mpsc::channel::<P::Block>(64);
             let block_provider = self.block_provider.clone();
@@ -99,7 +95,7 @@ where
             let block_hash = payload.block_hash();
             let block_number = payload.block_number();
 
-            previous_block_hashes.push(block_hash);
+            previous_block_hashes.enqueue(block_hash);
 
             // Send new events to execution client
             let _ = self.engine_handle.new_payload(payload).await;
@@ -142,3 +138,60 @@ where
         }
     }
 }
+
+/// Looks up a block hash from the ring buffer at the given offset from the most recent entry.
+///
+/// Returns `None` if the buffer doesn't have enough entries to satisfy the offset.
+fn get_hash_at_offset(buffer: &AllocRingBuffer<B256>, offset: usize) -> Option<B256> {
+    buffer.len().checked_sub(offset + 1).and_then(|index| buffer.get(index).copied())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_get_hash_at_offset() {
+        let mut buffer: AllocRingBuffer<B256> = AllocRingBuffer::new(65);
+
+        // Empty buffer returns None for any offset
+        assert_eq!(get_hash_at_offset(&buffer, 0), None);
+        assert_eq!(get_hash_at_offset(&buffer, 1), None);
+
+        // Push hashes 0..65
+        for i in 0..65u8 {
+            buffer.enqueue(B256::with_last_byte(i));
+        }
+
+        // offset=0 should return the most recent (64)
+        assert_eq!(get_hash_at_offset(&buffer, 0), Some(B256::with_last_byte(64)));
+
+        // offset=32 (safe block) should return hash 32
+        assert_eq!(get_hash_at_offset(&buffer, 32), Some(B256::with_last_byte(32)));
+
+        // offset=64 (finalized block) should return hash 0 (the oldest)
+        assert_eq!(get_hash_at_offset(&buffer, 64), Some(B256::with_last_byte(0)));
+
+        // offset=65 exceeds buffer, should return None
+        assert_eq!(get_hash_at_offset(&buffer, 65), None);
+    }
+
+    #[test]
+    fn test_get_hash_at_offset_insufficient_entries() {
+        let mut buffer: AllocRingBuffer<B256> = AllocRingBuffer::new(65);
+
+        // With only 1 entry, only offset=0 works
+        buffer.enqueue(B256::with_last_byte(1));
+        assert_eq!(get_hash_at_offset(&buffer, 0), Some(B256::with_last_byte(1)));
+        assert_eq!(get_hash_at_offset(&buffer, 1), None);
+        assert_eq!(get_hash_at_offset(&buffer, 32), None);
+        assert_eq!(get_hash_at_offset(&buffer, 64), None);
+
+        // With 33 entries, offset=32 works but offset=64 doesn't
+        for i in 2..=33u8 {
+            buffer.enqueue(B256::with_last_byte(i));
+        }
+        assert_eq!(get_hash_at_offset(&buffer, 32), Some(B256::with_last_byte(1)));
+        assert_eq!(get_hash_at_offset(&buffer, 64), None);
+    }
+}

crates/e2e-test-utils/src/setup_import.rs

@@ -114,19 +114,22 @@ pub async fn setup_engine_with_chain_import(
 
         // Initialize the database using init_db (same as CLI import command)
         let db_args = reth_node_core::args::DatabaseArgs::default().database_args();
-        let db_env = reth_db::init_db(&db_path, db_args)?;
-        let db = Arc::new(db_env);
+        let db = reth_db::init_db(&db_path, db_args)?;
 
         // Create a provider factory with the initialized database (use regular DB, not
         // TempDatabase) We need to specify the node types properly for the adapter
-        let provider_factory = ProviderFactory::<
-            NodeTypesWithDBAdapter<EthereumNode, Arc<DatabaseEnv>>,
-        >::new(
-            db.clone(),
-            chain_spec.clone(),
-            reth_provider::providers::StaticFileProvider::read_write(static_files_path.clone())?,
-            reth_provider::providers::RocksDBProvider::builder(rocksdb_dir_path).build().unwrap(),
-        )?;
+        let provider_factory =
+            ProviderFactory::<NodeTypesWithDBAdapter<EthereumNode, DatabaseEnv>>::new(
+                db.clone(),
+                chain_spec.clone(),
+                reth_provider::providers::StaticFileProvider::read_write(
+                    static_files_path.clone(),
+                )?,
+                reth_provider::providers::RocksDBProvider::builder(rocksdb_dir_path)
+                    .with_default_tables()
+                    .build()
+                    .unwrap(),
+            )?;
 
         // Initialize genesis if needed
         reth_db_common::init::init_genesis(&provider_factory)?;
@@ -317,17 +320,17 @@ mod tests {
         // Import the chain
         {
             let db_args = reth_node_core::args::DatabaseArgs::default().database_args();
-            let db_env = reth_db::init_db(&db_path, db_args).unwrap();
-            let db = Arc::new(db_env);
+            let db = reth_db::init_db(&db_path, db_args).unwrap();
 
             let provider_factory: ProviderFactory<
-                NodeTypesWithDBAdapter<reth_node_ethereum::EthereumNode, Arc<DatabaseEnv>>,
+                NodeTypesWithDBAdapter<reth_node_ethereum::EthereumNode, DatabaseEnv>,
             > = ProviderFactory::new(
                 db.clone(),
                 chain_spec.clone(),
                 reth_provider::providers::StaticFileProvider::read_write(static_files_path.clone())
                     .unwrap(),
                 reth_provider::providers::RocksDBProvider::builder(rocksdb_dir_path.clone())
+                    .with_default_tables()
                     .build()
                     .unwrap(),
             )
@@ -381,17 +384,17 @@ mod tests {
 
         // Now reopen the database and verify checkpoints are still there
         {
-            let db_env = reth_db::init_db(&db_path, DatabaseArguments::default()).unwrap();
-            let db = Arc::new(db_env);
+            let db = reth_db::init_db(&db_path, DatabaseArguments::default()).unwrap();
 
             let provider_factory: ProviderFactory<
-                NodeTypesWithDBAdapter<reth_node_ethereum::EthereumNode, Arc<DatabaseEnv>>,
+                NodeTypesWithDBAdapter<reth_node_ethereum::EthereumNode, DatabaseEnv>,
             > = ProviderFactory::new(
                 db,
                 chain_spec.clone(),
                 reth_provider::providers::StaticFileProvider::read_only(static_files_path, false)
                     .unwrap(),
                 reth_provider::providers::RocksDBProvider::builder(rocksdb_dir_path)
+                    .with_default_tables()
                     .build()
                     .unwrap(),
             )
@@ -490,7 +493,10 @@ mod tests {
             db.clone(),
             chain_spec.clone(),
             reth_provider::providers::StaticFileProvider::read_write(static_files_path).unwrap(),
-            reth_provider::providers::RocksDBProvider::builder(rocksdb_dir_path).build().unwrap(),
+            reth_provider::providers::RocksDBProvider::builder(rocksdb_dir_path)
+                .with_default_tables()
+                .build()
+                .unwrap(),
         )
         .expect("failed to create provider factory");
 

crates/e2e-test-utils/tests/rocksdb/main.rs

@@ -10,11 +10,10 @@ use jsonrpsee::core::client::ClientT;
 use reth_chainspec::{ChainSpec, ChainSpecBuilder, MAINNET};
 use reth_db::tables;
 use reth_e2e_test_utils::{transaction::TransactionTestContext, wallet, E2ETestSetupBuilder};
-use reth_node_builder::NodeConfig;
 use reth_node_core::args::RocksDbArgs;
 use reth_node_ethereum::EthereumNode;
 use reth_payload_builder::EthPayloadBuilderAttributes;
-use reth_provider::RocksDBProviderFactory;
+use reth_provider::{RocksDBProviderFactory, StorageSettings};
 use std::{sync::Arc, time::Duration};
 
 const ROCKSDB_POLL_TIMEOUT: Duration = Duration::from_secs(60);
@@ -97,16 +96,24 @@ fn test_attributes_generator(timestamp: u64) -> EthPayloadBuilderAttributes {
     EthPayloadBuilderAttributes::new(B256::ZERO, attributes)
 }
 
-/// Enables `RocksDB` for `TransactionHashNumbers` table.
-///
-/// Note: Static file changesets are disabled because `persistence_threshold(0)` causes
-/// a race where the static file writer expects sequential block numbers but receives
-/// them out of order, resulting in `UnexpectedStaticFileBlockNumber` errors.
-fn with_rocksdb_enabled<C>(mut config: NodeConfig<C>) -> NodeConfig<C> {
-    config.rocksdb = RocksDbArgs { tx_hash: true, ..Default::default() };
-    config.static_files.storage_changesets = false;
-    config.static_files.account_changesets = false;
-    config
+/// Verifies that `RocksDB` CLI defaults match `StorageSettings::base()`.
+#[test]
+fn test_rocksdb_defaults_match_storage_settings() {
+    let args = RocksDbArgs::default();
+    let settings = StorageSettings::base();
+
+    assert_eq!(
+        args.tx_hash, settings.transaction_hash_numbers_in_rocksdb,
+        "tx_hash default should match StorageSettings::base()"
+    );
+    assert_eq!(
+        args.storages_history, settings.storages_history_in_rocksdb,
+        "storages_history default should match StorageSettings::base()"
+    );
+    assert_eq!(
+        args.account_history, settings.account_history_in_rocksdb,
+        "account_history default should match StorageSettings::base()"
+    );
 }
 
 /// Smoke test: node boots with `RocksDB` routing enabled.
@@ -118,7 +125,6 @@ async fn test_rocksdb_node_startup() -> Result<()> {
 
     let (nodes, _tasks, _wallet) =
         E2ETestSetupBuilder::<EthereumNode, _>::new(1, chain_spec, test_attributes_generator)
-            .with_node_config_modifier(with_rocksdb_enabled)
             .build()
             .await?;
 
@@ -146,7 +152,6 @@ async fn test_rocksdb_block_mining() -> Result<()> {
 
     let (mut nodes, _tasks, _wallet) =
         E2ETestSetupBuilder::<EthereumNode, _>::new(1, chain_spec, test_attributes_generator)
-            .with_node_config_modifier(with_rocksdb_enabled)
             .build()
             .await?;
 
@@ -203,7 +208,6 @@ async fn test_rocksdb_transaction_queries() -> Result<()> {
         chain_spec.clone(),
         test_attributes_generator,
     )
-    .with_node_config_modifier(with_rocksdb_enabled)
     .with_tree_config_modifier(|config| config.with_persistence_threshold(0))
     .build()
     .await?;
@@ -270,7 +274,6 @@ async fn test_rocksdb_multi_tx_same_block() -> Result<()> {
         chain_spec.clone(),
         test_attributes_generator,
     )
-    .with_node_config_modifier(with_rocksdb_enabled)
     .with_tree_config_modifier(|config| config.with_persistence_threshold(0))
     .build()
     .await?;
@@ -338,7 +341,6 @@ async fn test_rocksdb_txs_across_blocks() -> Result<()> {
         chain_spec.clone(),
         test_attributes_generator,
     )
-    .with_node_config_modifier(with_rocksdb_enabled)
     .with_tree_config_modifier(|config| config.with_persistence_threshold(0))
     .build()
     .await?;
@@ -423,7 +425,6 @@ async fn test_rocksdb_pending_tx_not_in_storage() -> Result<()> {
         chain_spec.clone(),
         test_attributes_generator,
     )
-    .with_node_config_modifier(with_rocksdb_enabled)
     .with_tree_config_modifier(|config| config.with_persistence_threshold(0))
     .build()
     .await?;
@@ -468,3 +469,123 @@ async fn test_rocksdb_pending_tx_not_in_storage() -> Result<()> {
 
     Ok(())
 }
+
+/// Reorg with `RocksDB`: verifies that unwind correctly reads changesets from
+/// storage-aware locations (static files vs MDBX) rather than directly from MDBX.
+///
+/// This test exercises `unwind_trie_state_from` which previously failed with
+/// `UnsortedInput` errors because it read changesets directly from MDBX tables
+/// instead of using storage-aware methods that check `storage_changesets_in_static_files`.
+#[tokio::test]
+async fn test_rocksdb_reorg_unwind() -> Result<()> {
+    reth_tracing::init_test_tracing();
+
+    let chain_spec = test_chain_spec();
+    let chain_id = chain_spec.chain().id();
+
+    let (mut nodes, _tasks, _) = E2ETestSetupBuilder::<EthereumNode, _>::new(
+        1,
+        chain_spec.clone(),
+        test_attributes_generator,
+    )
+    .with_tree_config_modifier(|config| config.with_persistence_threshold(0))
+    .build()
+    .await?;
+
+    assert_eq!(nodes.len(), 1);
+
+    // Use two separate wallets to avoid nonce conflicts during reorg
+    let wallets = wallet::Wallet::new(2).with_chain_id(chain_id).wallet_gen();
+    let signer1 = wallets[0].clone();
+    let signer2 = wallets[1].clone();
+    let client = nodes[0].rpc_client().expect("RPC client");
+
+    // Mine block 1 with a transaction from signer1
+    let raw_tx1 =
+        TransactionTestContext::transfer_tx_bytes_with_nonce(chain_id, signer1.clone(), 0).await;
+    let tx_hash1 = nodes[0].rpc.inject_tx(raw_tx1).await?;
+    wait_for_pending_tx(&client, tx_hash1).await;
+
+    let payload1 = nodes[0].advance_block().await?;
+    let block1_hash = payload1.block().hash();
+    assert_eq!(payload1.block().number(), 1);
+
+    // Poll until tx1 appears in RocksDB (ensures persistence happened)
+    let tx_number1 = poll_tx_in_rocksdb(&nodes[0].inner.provider, tx_hash1).await;
+    assert_eq!(tx_number1, 0, "First tx should have tx_number 0");
+
+    // Mine block 2 with transaction from signer1 (nonce 1)
+    let raw_tx2 =
+        TransactionTestContext::transfer_tx_bytes_with_nonce(chain_id, signer1.clone(), 1).await;
+    let tx_hash2 = nodes[0].rpc.inject_tx(raw_tx2).await?;
+    wait_for_pending_tx(&client, tx_hash2).await;
+
+    let payload2 = nodes[0].advance_block().await?;
+    assert_eq!(payload2.block().number(), 2);
+
+    // Poll until tx2 appears in RocksDB
+    let tx_number2 = poll_tx_in_rocksdb(&nodes[0].inner.provider, tx_hash2).await;
+    assert_eq!(tx_number2, 1, "Second tx should have tx_number 1");
+
+    // Mine block 3 with transaction from signer1 (nonce 2)
+    let raw_tx3 =
+        TransactionTestContext::transfer_tx_bytes_with_nonce(chain_id, signer1.clone(), 2).await;
+    let tx_hash3 = nodes[0].rpc.inject_tx(raw_tx3).await?;
+    wait_for_pending_tx(&client, tx_hash3).await;
+
+    let payload3 = nodes[0].advance_block().await?;
+    assert_eq!(payload3.block().number(), 3);
+
+    // Poll until tx3 appears in RocksDB
+    let tx_number3 = poll_tx_in_rocksdb(&nodes[0].inner.provider, tx_hash3).await;
+    assert_eq!(tx_number3, 2, "Third tx should have tx_number 2");
+
+    // Now create an alternate block 2 using signer2 (different wallet, avoids nonce conflict)
+    // Inject a tx from signer2 (nonce 0) before building the alternate block
+    let raw_alt_tx =
+        TransactionTestContext::transfer_tx_bytes_with_nonce(chain_id, signer2.clone(), 0).await;
+    let alt_tx_hash = nodes[0].rpc.inject_tx(raw_alt_tx).await?;
+    wait_for_pending_tx(&client, alt_tx_hash).await;
+
+    // Build an alternate payload (this builds on top of the current head, i.e., block 3)
+    // But we want to reorg back to block 1, so we'll use the payload and then FCU to it
+    let alt_payload = nodes[0].new_payload().await?;
+    let alt_block_hash = nodes[0].submit_payload(alt_payload.clone()).await?;
+
+    // Trigger reorg: make the alternate chain canonical by sending FCU pointing to block 1's hash
+    // as finalized, which should trigger an unwind of blocks 2 and 3
+    // The alt block becomes the new head
+    nodes[0].update_forkchoice(block1_hash, alt_block_hash).await?;
+
+    // Give time for the reorg to complete
+    tokio::time::sleep(Duration::from_millis(500)).await;
+
+    // Verify we can still query transactions and the chain is consistent
+    // If unwind_trie_state_from failed, this would have errored during reorg
+    let latest: Option<alloy_rpc_types_eth::Block> =
+        client.request("eth_getBlockByNumber", ("latest", false)).await?;
+    let latest = latest.expect("Latest block should exist");
+    // The alt block is at height 4 (on top of block 3)
+    assert!(latest.header.number >= 3, "Should be at height >= 3 after operation");
+
+    // tx1 from block 1 should still be there
+    let tx1: Option<Transaction> = client.request("eth_getTransactionByHash", [tx_hash1]).await?;
+    assert!(tx1.is_some(), "tx1 from block 1 should still be queryable");
+    assert_eq!(tx1.unwrap().block_number, Some(1));
+
+    // Mine another block to verify the chain can continue
+    let raw_tx_final =
+        TransactionTestContext::transfer_tx_bytes_with_nonce(chain_id, signer2.clone(), 1).await;
+    let tx_hash_final = nodes[0].rpc.inject_tx(raw_tx_final).await?;
+    wait_for_pending_tx(&client, tx_hash_final).await;
+
+    let final_payload = nodes[0].advance_block().await?;
+    assert!(final_payload.block().number() > 3, "Should be able to mine block after reorg");
+
+    // Verify tx_final is included
+    let tx_final: Option<Transaction> =
+        client.request("eth_getTransactionByHash", [tx_hash_final]).await?;
+    assert!(tx_final.is_some(), "final tx should be in latest block");
+
+    Ok(())
+}

crates/engine/local/Cargo.toml

@@ -32,9 +32,7 @@ futures-util.workspace = true
 # misc
 eyre.workspace = true
 tracing.workspace = true
-
 op-alloy-rpc-types-engine = { workspace = true, optional = true }
-reth-optimism-chainspec = { workspace = true, optional = true }
 
 [lints]
 workspace = true
@@ -42,7 +40,6 @@ workspace = true
 [features]
 op = [
     "dep:op-alloy-rpc-types-engine",
-    "dep:reth-optimism-chainspec",
     "reth-payload-primitives/op",
     "reth-primitives-traits/op",
 ]

crates/engine/local/src/payload.rs

@@ -72,13 +72,18 @@ where
         &self,
         parent: &SealedHeader<ChainSpec::Header>,
     ) -> op_alloy_rpc_types_engine::OpPayloadAttributes {
+        /// Dummy system transaction for dev mode.
+        /// OP Mainnet transaction at index 0 in block 124665056.
+        ///
+        /// <https://optimistic.etherscan.io/tx/0x312e290cf36df704a2217b015d6455396830b0ce678b860ebfcc30f41403d7b1>
+        const TX_SET_L1_BLOCK_OP_MAINNET_BLOCK_124665056: [u8; 251] = alloy_primitives::hex!(
+            "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"
+        );
+
         op_alloy_rpc_types_engine::OpPayloadAttributes {
             payload_attributes: self.build(parent),
             // Add dummy system transaction
-            transactions: Some(vec![
-                reth_optimism_chainspec::constants::TX_SET_L1_BLOCK_OP_MAINNET_BLOCK_124665056
-                    .into(),
-            ]),
+            transactions: Some(vec![TX_SET_L1_BLOCK_OP_MAINNET_BLOCK_124665056.into()]),
             no_tx_pool: None,
             gas_limit: None,
             eip_1559_params: None,

crates/engine/primitives/src/config.rs

@@ -47,6 +47,17 @@ pub const DEFAULT_RESERVED_CPU_CORES: usize = 1;
 /// Default maximum concurrency for prewarm task.
 pub const DEFAULT_PREWARM_MAX_CONCURRENCY: usize = 16;
 
+/// Default depth for sparse trie pruning.
+///
+/// Nodes at this depth and below are converted to hash stubs to reduce memory.
+/// Depth 4 means we keep roughly 16^4 = 65536 potential branch paths at most.
+pub const DEFAULT_SPARSE_TRIE_PRUNE_DEPTH: usize = 4;
+
+/// Default maximum number of storage tries to keep after pruning.
+///
+/// Storage tries beyond this limit are cleared (but allocations preserved).
+pub const DEFAULT_SPARSE_TRIE_MAX_STORAGE_TRIES: usize = 100;
+
 const DEFAULT_BLOCK_BUFFER_LIMIT: u32 = EPOCH_SLOTS as u32 * 2;
 const DEFAULT_MAX_INVALID_HEADER_CACHE_LENGTH: u32 = 256;
 const DEFAULT_MAX_EXECUTE_BLOCK_BATCH_SIZE: usize = 4;
@@ -148,10 +159,16 @@ pub struct TreeConfig {
     storage_worker_count: usize,
     /// Number of account proof worker threads.
     account_worker_count: usize,
-    /// Whether to enable V2 storage proofs.
-    enable_proof_v2: bool,
+    /// Whether to disable V2 storage proofs.
+    disable_proof_v2: bool,
     /// Whether to disable cache metrics recording (can be expensive with large cached state).
     disable_cache_metrics: bool,
+    /// Whether to enable sparse trie as cache.
+    enable_sparse_trie_as_cache: bool,
+    /// Depth for sparse trie pruning after state root computation.
+    sparse_trie_prune_depth: usize,
+    /// Maximum number of storage tries to retain after pruning.
+    sparse_trie_max_storage_tries: usize,
 }
 
 impl Default for TreeConfig {
@@ -179,8 +196,11 @@ impl Default for TreeConfig {
             allow_unwind_canonical_header: false,
             storage_worker_count: default_storage_worker_count(),
             account_worker_count: default_account_worker_count(),
-            enable_proof_v2: false,
+            disable_proof_v2: false,
             disable_cache_metrics: false,
+            enable_sparse_trie_as_cache: false,
+            sparse_trie_prune_depth: DEFAULT_SPARSE_TRIE_PRUNE_DEPTH,
+            sparse_trie_max_storage_tries: DEFAULT_SPARSE_TRIE_MAX_STORAGE_TRIES,
         }
     }
 }
@@ -211,8 +231,10 @@ impl TreeConfig {
         allow_unwind_canonical_header: bool,
         storage_worker_count: usize,
         account_worker_count: usize,
-        enable_proof_v2: bool,
+        disable_proof_v2: bool,
         disable_cache_metrics: bool,
+        sparse_trie_prune_depth: usize,
+        sparse_trie_max_storage_tries: usize,
     ) -> Self {
         Self {
             persistence_threshold,
@@ -237,8 +259,11 @@ impl TreeConfig {
             allow_unwind_canonical_header,
             storage_worker_count,
             account_worker_count,
-            enable_proof_v2,
+            disable_proof_v2,
             disable_cache_metrics,
+            enable_sparse_trie_as_cache: false,
+            sparse_trie_prune_depth,
+            sparse_trie_max_storage_tries,
         }
     }
 
@@ -280,7 +305,8 @@ impl TreeConfig {
     /// Return the multiproof task chunk size, using the V2 default if V2 proofs are enabled
     /// and the chunk size is at the default value.
     pub const fn effective_multiproof_chunk_size(&self) -> usize {
-        if self.enable_proof_v2 && self.multiproof_chunk_size == DEFAULT_MULTIPROOF_TASK_CHUNK_SIZE
+        if !self.disable_proof_v2 &&
+            self.multiproof_chunk_size == DEFAULT_MULTIPROOF_TASK_CHUNK_SIZE
         {
             DEFAULT_MULTIPROOF_TASK_CHUNK_SIZE_V2
         } else {
@@ -502,8 +528,12 @@ impl TreeConfig {
     }
 
     /// Setter for the number of storage proof worker threads.
-    pub fn with_storage_worker_count(mut self, storage_worker_count: usize) -> Self {
-        self.storage_worker_count = storage_worker_count.max(MIN_WORKER_COUNT);
+    ///
+    /// No-op if it's [`None`].
+    pub fn with_storage_worker_count_opt(mut self, storage_worker_count: Option<usize>) -> Self {
+        if let Some(count) = storage_worker_count {
+            self.storage_worker_count = count.max(MIN_WORKER_COUNT);
+        }
         self
     }
 
@@ -513,19 +543,23 @@ impl TreeConfig {
     }
 
     /// Setter for the number of account proof worker threads.
-    pub fn with_account_worker_count(mut self, account_worker_count: usize) -> Self {
-        self.account_worker_count = account_worker_count.max(MIN_WORKER_COUNT);
+    ///
+    /// No-op if it's [`None`].
+    pub fn with_account_worker_count_opt(mut self, account_worker_count: Option<usize>) -> Self {
+        if let Some(count) = account_worker_count {
+            self.account_worker_count = count.max(MIN_WORKER_COUNT);
+        }
         self
     }
 
-    /// Return whether V2 storage proofs are enabled.
-    pub const fn enable_proof_v2(&self) -> bool {
-        self.enable_proof_v2
+    /// Return whether V2 storage proofs are disabled.
+    pub const fn disable_proof_v2(&self) -> bool {
+        self.disable_proof_v2
     }
 
-    /// Setter for whether to enable V2 storage proofs.
-    pub const fn with_enable_proof_v2(mut self, enable_proof_v2: bool) -> Self {
-        self.enable_proof_v2 = enable_proof_v2;
+    /// Setter for whether to disable V2 storage proofs.
+    pub const fn with_disable_proof_v2(mut self, disable_proof_v2: bool) -> Self {
+        self.disable_proof_v2 = disable_proof_v2;
         self
     }
 
@@ -539,4 +573,37 @@ impl TreeConfig {
         self.disable_cache_metrics = disable_cache_metrics;
         self
     }
+
+    /// Returns whether sparse trie as cache is enabled.
+    pub const fn enable_sparse_trie_as_cache(&self) -> bool {
+        self.enable_sparse_trie_as_cache
+    }
+
+    /// Setter for whether to enable sparse trie as cache.
+    pub const fn with_enable_sparse_trie_as_cache(mut self, value: bool) -> Self {
+        self.enable_sparse_trie_as_cache = value;
+        self
+    }
+
+    /// Returns the sparse trie prune depth.
+    pub const fn sparse_trie_prune_depth(&self) -> usize {
+        self.sparse_trie_prune_depth
+    }
+
+    /// Setter for sparse trie prune depth.
+    pub const fn with_sparse_trie_prune_depth(mut self, depth: usize) -> Self {
+        self.sparse_trie_prune_depth = depth;
+        self
+    }
+
+    /// Returns the maximum number of storage tries to retain after pruning.
+    pub const fn sparse_trie_max_storage_tries(&self) -> usize {
+        self.sparse_trie_max_storage_tries
+    }
+
+    /// Setter for maximum storage tries to retain.
+    pub const fn with_sparse_trie_max_storage_tries(mut self, max_tries: usize) -> Self {
+        self.sparse_trie_max_storage_tries = max_tries;
+        self
+    }
 }

crates/engine/service/Cargo.toml

@@ -17,7 +17,6 @@ reth-engine-tree.workspace = true
 reth-evm.workspace = true
 reth-network-p2p.workspace = true
 reth-payload-builder.workspace = true
-reth-ethereum-primitives.workspace = true
 reth-provider.workspace = true
 reth-prune.workspace = true
 reth-stages-api.workspace = true

crates/engine/service/src/service.rs

@@ -14,7 +14,6 @@ pub use reth_engine_tree::{
     chain::{ChainEvent, ChainOrchestrator},
     engine::EngineApiEvent,
 };
-use reth_ethereum_primitives::EthPrimitives;
 use reth_evm::ConfigureEvm;
 use reth_network_p2p::BlockClient;
 use reth_node_types::{BlockTy, NodeTypes};
@@ -97,11 +96,11 @@ where
         let downloader = BasicBlockDownloader::new(client, consensus.clone());
 
         let persistence_handle =
-            PersistenceHandle::<EthPrimitives>::spawn_service(provider, pruner, sync_metrics_tx);
+            PersistenceHandle::<N::Primitives>::spawn_service(provider, pruner, sync_metrics_tx);
 
         let canonical_in_memory_state = blockchain_db.canonical_in_memory_state();
 
-        let (to_tree_tx, from_tree) = EngineApiTreeHandler::<N::Primitives, _, _, _, _>::spawn_new(
+        let (to_tree_tx, from_tree) = EngineApiTreeHandler::spawn_new(
             blockchain_db,
             consensus,
             payload_validator,

crates/engine/tree/Cargo.toml

@@ -23,7 +23,7 @@ reth-evm = { workspace = true, features = ["metrics"] }
 reth-network-p2p.workspace = true
 reth-payload-builder.workspace = true
 reth-payload-primitives.workspace = true
-reth-primitives-traits.workspace = true
+reth-primitives-traits = { workspace = true, features = ["rayon"] }
 reth-ethereum-primitives.workspace = true
 reth-provider.workspace = true
 reth-prune.workspace = true

crates/engine/tree/src/metrics.rs

@@ -20,7 +20,7 @@ pub(crate) struct PersistenceMetrics {
     /// How long it took for blocks to be saved
     pub(crate) save_blocks_duration_seconds: Histogram,
     /// How many blocks we persist at once.
-    pub(crate) save_blocks_block_count: Histogram,
+    pub(crate) save_blocks_batch_size: Histogram,
     /// How long it took for blocks to be pruned
     pub(crate) prune_before_duration_seconds: Histogram,
 }

crates/engine/tree/src/persistence.rs

@@ -12,7 +12,11 @@ use reth_provider::{
 use reth_prune::{PrunerError, PrunerOutput, PrunerWithFactory};
 use reth_stages_api::{MetricEvent, MetricEventsSender};
 use std::{
-    sync::mpsc::{Receiver, SendError, Sender},
+    sync::{
+        mpsc::{Receiver, SendError, Sender},
+        Arc,
+    },
+    thread::JoinHandle,
     time::Instant,
 };
 use thiserror::Error;
@@ -40,6 +44,12 @@ where
     metrics: PersistenceMetrics,
     /// Sender for sync metrics - we only submit sync metrics for persisted blocks
     sync_metrics_tx: MetricEventsSender,
+    /// Pending finalized block number to be committed with the next block save.
+    /// This avoids triggering a separate fsync for each finalized block update.
+    pending_finalized_block: Option<u64>,
+    /// Pending safe block number to be committed with the next block save.
+    /// This avoids triggering a separate fsync for each safe block update.
+    pending_safe_block: Option<u64>,
 }
 
 impl<N> PersistenceService<N>
@@ -53,7 +63,15 @@ where
         pruner: PrunerWithFactory<ProviderFactory<N>>,
         sync_metrics_tx: MetricEventsSender,
     ) -> Self {
-        Self { provider, incoming, pruner, metrics: PersistenceMetrics::default(), sync_metrics_tx }
+        Self {
+            provider,
+            incoming,
+            pruner,
+            metrics: PersistenceMetrics::default(),
+            sync_metrics_tx,
+            pending_finalized_block: None,
+            pending_safe_block: None,
+        }
     }
 
     /// Prunes block data before the given block number according to the configured prune
@@ -106,14 +124,10 @@ where
                     }
                 }
                 PersistenceAction::SaveFinalizedBlock(finalized_block) => {
-                    let provider = self.provider.database_provider_rw()?;
-                    provider.save_finalized_block_number(finalized_block)?;
-                    provider.commit()?;
+                    self.pending_finalized_block = Some(finalized_block);
                 }
                 PersistenceAction::SaveSafeBlock(safe_block) => {
-                    let provider = self.provider.database_provider_rw()?;
-                    provider.save_safe_block_number(safe_block)?;
-                    provider.commit()?;
+                    self.pending_safe_block = Some(safe_block);
                 }
             }
         }
@@ -138,26 +152,39 @@ where
     }
 
     fn on_save_blocks(
-        &self,
+        &mut self,
         blocks: Vec<ExecutedBlock<N::Primitives>>,
     ) -> Result<Option<BlockNumHash>, PersistenceError> {
         let first_block = blocks.first().map(|b| b.recovered_block.num_hash());
         let last_block = blocks.last().map(|b| b.recovered_block.num_hash());
         let block_count = blocks.len();
+
+        // Take any pending finalized/safe block updates to commit together
+        let pending_finalized = self.pending_finalized_block.take();
+        let pending_safe = self.pending_safe_block.take();
+
         debug!(target: "engine::persistence", ?block_count, first=?first_block, last=?last_block, "Saving range of blocks");
 
         let start_time = Instant::now();
 
         if last_block.is_some() {
             let provider_rw = self.provider.database_provider_rw()?;
-
             provider_rw.save_blocks(blocks, SaveBlocksMode::Full)?;
+
+            // Commit pending finalized/safe block updates in the same transaction
+            if let Some(finalized) = pending_finalized {
+                provider_rw.save_finalized_block_number(finalized)?;
+            }
+            if let Some(safe) = pending_safe {
+                provider_rw.save_safe_block_number(safe)?;
+            }
+
             provider_rw.commit()?;
         }
 
         debug!(target: "engine::persistence", first=?first_block, last=?last_block, "Saved range of blocks");
 
-        self.metrics.save_blocks_block_count.record(block_count as f64);
+        self.metrics.save_blocks_batch_size.record(block_count as f64);
         self.metrics.save_blocks_duration_seconds.record(start_time.elapsed());
 
         Ok(last_block)
@@ -204,15 +231,25 @@ pub enum PersistenceAction<N: NodePrimitives = EthPrimitives> {
 pub struct PersistenceHandle<N: NodePrimitives = EthPrimitives> {
     /// The channel used to communicate with the persistence service
     sender: Sender<PersistenceAction<N>>,
+    /// Guard that joins the service thread when all handles are dropped.
+    /// Uses `Arc` so the handle remains `Clone`.
+    _service_guard: Arc<ServiceGuard>,
 }
 
 impl<T: NodePrimitives> PersistenceHandle<T> {
     /// Create a new [`PersistenceHandle`] from a [`Sender<PersistenceAction>`].
-    pub const fn new(sender: Sender<PersistenceAction<T>>) -> Self {
-        Self { sender }
+    ///
+    /// This is intended for testing purposes where you want to mock the persistence service.
+    /// For production use, prefer [`spawn_service`](Self::spawn_service).
+    pub fn new(sender: Sender<PersistenceAction<T>>) -> Self {
+        Self { sender, _service_guard: Arc::new(ServiceGuard(None)) }
     }
 
     /// Create a new [`PersistenceHandle`], and spawn the persistence service.
+    ///
+    /// The returned handle can be cloned and shared. When all clones are dropped, the service
+    /// thread will be joined, ensuring graceful shutdown before resources (like `RocksDB`) are
+    /// released.
     pub fn spawn_service<N>(
         provider_factory: ProviderFactory<N>,
         pruner: PrunerWithFactory<ProviderFactory<N>>,
@@ -224,13 +261,10 @@ impl<T: NodePrimitives> PersistenceHandle<T> {
         // create the initial channels
         let (db_service_tx, db_service_rx) = std::sync::mpsc::channel();
 
-        // construct persistence handle
-        let persistence_handle = PersistenceHandle::new(db_service_tx);
-
         // spawn the persistence service
         let db_service =
             PersistenceService::new(provider_factory, db_service_rx, pruner, sync_metrics_tx);
-        std::thread::Builder::new()
+        let join_handle = std::thread::Builder::new()
             .name("Persistence Service".to_string())
             .spawn(|| {
                 if let Err(err) = db_service.run() {
@@ -239,7 +273,10 @@ impl<T: NodePrimitives> PersistenceHandle<T> {
             })
             .unwrap();
 
-        persistence_handle
+        PersistenceHandle {
+            sender: db_service_tx,
+            _service_guard: Arc::new(ServiceGuard(Some(join_handle))),
+        }
     }
 
     /// Sends a specific [`PersistenceAction`] in the contained channel. The caller is responsible
@@ -267,7 +304,10 @@ impl<T: NodePrimitives> PersistenceHandle<T> {
         self.send_action(PersistenceAction::SaveBlocks(blocks, tx))
     }
 
-    /// Persists the finalized block number on disk.
+    /// Queues the finalized block number to be persisted on disk.
+    ///
+    /// The update is deferred and will be committed together with the next [`Self::save_blocks`]
+    /// call to avoid triggering a separate fsync for each update.
     pub fn save_finalized_block_number(
         &self,
         finalized_block: u64,
@@ -275,7 +315,10 @@ impl<T: NodePrimitives> PersistenceHandle<T> {
         self.send_action(PersistenceAction::SaveFinalizedBlock(finalized_block))
     }
 
-    /// Persists the safe block number on disk.
+    /// Queues the safe block number to be persisted on disk.
+    ///
+    /// The update is deferred and will be committed together with the next [`Self::save_blocks`]
+    /// call to avoid triggering a separate fsync for each update.
     pub fn save_safe_block_number(
         &self,
         safe_block: u64,
@@ -297,6 +340,27 @@ impl<T: NodePrimitives> PersistenceHandle<T> {
     }
 }
 
+/// Guard that joins the persistence service thread when dropped.
+///
+/// This ensures graceful shutdown - the service thread completes before resources like
+/// `RocksDB` are released. Stored in an `Arc` inside [`PersistenceHandle`] so the handle
+/// can be cloned while sharing the same guard.
+struct ServiceGuard(Option<JoinHandle<()>>);
+
+impl std::fmt::Debug for ServiceGuard {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_tuple("ServiceGuard").field(&self.0.as_ref().map(|_| "...")).finish()
+    }
+}
+
+impl Drop for ServiceGuard {
+    fn drop(&mut self) {
+        if let Some(join_handle) = self.0.take() {
+            let _ = join_handle.join();
+        }
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -323,12 +387,12 @@ mod tests {
     #[test]
     fn test_save_blocks_empty() {
         reth_tracing::init_test_tracing();
-        let persistence_handle = default_persistence_handle();
+        let handle = default_persistence_handle();
 
         let blocks = vec![];
         let (tx, rx) = crossbeam_channel::bounded(1);
 
-        persistence_handle.save_blocks(blocks, tx).unwrap();
+        handle.save_blocks(blocks, tx).unwrap();
 
         let hash = rx.recv().unwrap();
         assert_eq!(hash, None);
@@ -337,7 +401,7 @@ mod tests {
     #[test]
     fn test_save_blocks_single_block() {
         reth_tracing::init_test_tracing();
-        let persistence_handle = default_persistence_handle();
+        let handle = default_persistence_handle();
         let block_number = 0;
         let mut test_block_builder = TestBlockBuilder::eth();
         let executed =
@@ -347,7 +411,7 @@ mod tests {
         let blocks = vec![executed];
         let (tx, rx) = crossbeam_channel::bounded(1);
 
-        persistence_handle.save_blocks(blocks, tx).unwrap();
+        handle.save_blocks(blocks, tx).unwrap();
 
         let BlockNumHash { hash: actual_hash, number: _ } = rx
             .recv_timeout(std::time::Duration::from_secs(10))
@@ -360,14 +424,14 @@ mod tests {
     #[test]
     fn test_save_blocks_multiple_blocks() {
         reth_tracing::init_test_tracing();
-        let persistence_handle = default_persistence_handle();
+        let handle = default_persistence_handle();
 
         let mut test_block_builder = TestBlockBuilder::eth();
         let blocks = test_block_builder.get_executed_blocks(0..5).collect::<Vec<_>>();
         let last_hash = blocks.last().unwrap().recovered_block().hash();
         let (tx, rx) = crossbeam_channel::bounded(1);
 
-        persistence_handle.save_blocks(blocks, tx).unwrap();
+        handle.save_blocks(blocks, tx).unwrap();
         let BlockNumHash { hash: actual_hash, number: _ } = rx.recv().unwrap().unwrap();
         assert_eq!(last_hash, actual_hash);
     }
@@ -375,7 +439,7 @@ mod tests {
     #[test]
     fn test_save_blocks_multiple_calls() {
         reth_tracing::init_test_tracing();
-        let persistence_handle = default_persistence_handle();
+        let handle = default_persistence_handle();
 
         let ranges = [0..1, 1..2, 2..4, 4..5];
         let mut test_block_builder = TestBlockBuilder::eth();
@@ -384,7 +448,7 @@ mod tests {
             let last_hash = blocks.last().unwrap().recovered_block().hash();
             let (tx, rx) = crossbeam_channel::bounded(1);
 
-            persistence_handle.save_blocks(blocks, tx).unwrap();
+            handle.save_blocks(blocks, tx).unwrap();
 
             let BlockNumHash { hash: actual_hash, number: _ } = rx.recv().unwrap().unwrap();
             assert_eq!(last_hash, actual_hash);

crates/engine/tree/src/tree/cached_state.rs

@@ -76,7 +76,8 @@ impl CacheConfig for EpochCacheConfig {
 type FixedCache<K, V, H = DefaultHashBuilder> = fixed_cache::Cache<K, V, H, EpochCacheConfig>;
 
 /// A wrapper of a state provider and a shared cache.
-pub(crate) struct CachedStateProvider<S> {
+#[derive(Debug)]
+pub struct CachedStateProvider<S> {
     /// The state provider
     state_provider: S,
 
@@ -96,7 +97,7 @@ where
 {
     /// Creates a new [`CachedStateProvider`] from an [`ExecutionCache`], state provider, and
     /// [`CachedStateMetrics`].
-    pub(crate) const fn new(
+    pub const fn new(
         state_provider: S,
         caches: ExecutionCache,
         metrics: CachedStateMetrics,
@@ -114,7 +115,7 @@ impl<S> CachedStateProvider<S> {
     /// [`State`](revm::database::State) also caches internally during block execution and the cache
     /// is then updated after the block with the entire [`BundleState`] output of that block which
     /// contains all accessed accounts,code,storage. See also [`ExecutionCache::insert_state`].
-    pub(crate) const fn prewarm(mut self) -> Self {
+    pub const fn prewarm(mut self) -> Self {
         self.prewarm = true;
         self
     }
@@ -131,7 +132,7 @@ impl<S> CachedStateProvider<S> {
 /// and the fixed-cache internal stats (collisions, size, capacity).
 #[derive(Metrics, Clone)]
 #[metrics(scope = "sync.caching")]
-pub(crate) struct CachedStateMetrics {
+pub struct CachedStateMetrics {
     /// Number of times a new execution cache was created
     execution_cache_created_total: Counter,
 
@@ -186,7 +187,7 @@ pub(crate) struct CachedStateMetrics {
 
 impl CachedStateMetrics {
     /// Sets all values to zero, indicating that a new block is being executed.
-    pub(crate) fn reset(&self) {
+    pub fn reset(&self) {
         // code cache
         self.code_cache_hits.set(0);
         self.code_cache_misses.set(0);
@@ -204,7 +205,7 @@ impl CachedStateMetrics {
     }
 
     /// Returns a new zeroed-out instance of [`CachedStateMetrics`].
-    pub(crate) fn zeroed() -> Self {
+    pub fn zeroed() -> Self {
         let zeroed = Self::default();
         zeroed.reset();
         zeroed
@@ -312,14 +313,7 @@ impl<S: AccountReader> AccountReader for CachedStateProvider<S> {
             match self.caches.get_or_try_insert_account_with(*address, || {
                 self.state_provider.basic_account(address)
             })? {
-                CachedStatus::NotCached(value) => {
-                    self.metrics.account_cache_misses.increment(1);
-                    Ok(value)
-                }
-                CachedStatus::Cached(value) => {
-                    self.metrics.account_cache_hits.increment(1);
-                    Ok(value)
-                }
+                CachedStatus::NotCached(value) | CachedStatus::Cached(value) => Ok(value),
             }
         } else if let Some(account) = self.caches.account_cache.get(address) {
             self.metrics.account_cache_hits.increment(1);
@@ -333,7 +327,7 @@ impl<S: AccountReader> AccountReader for CachedStateProvider<S> {
 
 /// Represents the status of a key in the cache.
 #[derive(Debug, Clone, PartialEq, Eq)]
-pub(crate) enum CachedStatus<T> {
+pub enum CachedStatus<T> {
     /// The key is not in the cache (or was invalidated). The value was recalculated.
     NotCached(T),
     /// The key exists in cache and has a specific value.
@@ -350,14 +344,7 @@ impl<S: StateProvider> StateProvider for CachedStateProvider<S> {
             match self.caches.get_or_try_insert_storage_with(account, storage_key, || {
                 self.state_provider.storage(account, storage_key).map(Option::unwrap_or_default)
             })? {
-                CachedStatus::NotCached(value) => {
-                    self.metrics.storage_cache_misses.increment(1);
-                    // The slot that was never written to is indistinguishable from a slot
-                    // explicitly set to zero. We return `None` in both cases.
-                    Ok(Some(value).filter(|v| !v.is_zero()))
-                }
-                CachedStatus::Cached(value) => {
-                    self.metrics.storage_cache_hits.increment(1);
+                CachedStatus::NotCached(value) | CachedStatus::Cached(value) => {
                     // The slot that was never written to is indistinguishable from a slot
                     // explicitly set to zero. We return `None` in both cases.
                     Ok(Some(value).filter(|v| !v.is_zero()))
@@ -379,14 +366,7 @@ impl<S: BytecodeReader> BytecodeReader for CachedStateProvider<S> {
             match self.caches.get_or_try_insert_code_with(*code_hash, || {
                 self.state_provider.bytecode_by_hash(code_hash)
             })? {
-                CachedStatus::NotCached(code) => {
-                    self.metrics.code_cache_misses.increment(1);
-                    Ok(code)
-                }
-                CachedStatus::Cached(code) => {
-                    self.metrics.code_cache_hits.increment(1);
-                    Ok(code)
-                }
+                CachedStatus::NotCached(code) | CachedStatus::Cached(code) => Ok(code),
             }
         } else if let Some(code) = self.caches.code_cache.get(code_hash) {
             self.metrics.code_cache_hits.increment(1);
@@ -508,7 +488,7 @@ impl<S: HashedPostStateProvider> HashedPostStateProvider for CachedStateProvider
 /// Since EIP-6780, SELFDESTRUCT only works within the same transaction where the
 /// contract was created, so we don't need to handle clearing the storage.
 #[derive(Debug, Clone)]
-pub(crate) struct ExecutionCache {
+pub struct ExecutionCache {
     /// Cache for contract bytecode, keyed by code hash.
     code_cache: Arc<FixedCache<B256, Option<Bytecode>, FbBuildHasher<32>>>,
 
@@ -540,7 +520,7 @@ impl ExecutionCache {
     ///
     /// Fixed-cache requires power-of-two sizes for efficient indexing.
     /// With epochs enabled, the minimum size is 4096 entries.
-    pub(crate) const fn bytes_to_entries(size_bytes: usize, entry_size: usize) -> usize {
+    pub const fn bytes_to_entries(size_bytes: usize, entry_size: usize) -> usize {
         let entries = size_bytes / entry_size;
         // Round down to nearest power of two
         let rounded = if entries == 0 { 1 } else { (entries + 1).next_power_of_two() >> 1 };
@@ -553,7 +533,7 @@ impl ExecutionCache {
     }
 
     /// Build an [`ExecutionCache`] struct, so that execution caches can be easily cloned.
-    pub(crate) fn new(total_cache_size: usize) -> Self {
+    pub fn new(total_cache_size: usize) -> Self {
         let storage_cache_size = (total_cache_size * 8888) / 10000; // 88.88% of total
         let account_cache_size = (total_cache_size * 556) / 10000; // 5.56% of total
         let code_cache_size = (total_cache_size * 556) / 10000; // 5.56% of total
@@ -587,7 +567,7 @@ impl ExecutionCache {
     }
 
     /// Gets code from cache, or inserts using the provided function.
-    pub(crate) fn get_or_try_insert_code_with<E>(
+    pub fn get_or_try_insert_code_with<E>(
         &self,
         hash: B256,
         f: impl FnOnce() -> Result<Option<Bytecode>, E>,
@@ -606,7 +586,7 @@ impl ExecutionCache {
     }
 
     /// Gets storage from cache, or inserts using the provided function.
-    pub(crate) fn get_or_try_insert_storage_with<E>(
+    pub fn get_or_try_insert_storage_with<E>(
         &self,
         address: Address,
         key: StorageKey,
@@ -626,7 +606,7 @@ impl ExecutionCache {
     }
 
     /// Gets account from cache, or inserts using the provided function.
-    pub(crate) fn get_or_try_insert_account_with<E>(
+    pub fn get_or_try_insert_account_with<E>(
         &self,
         address: Address,
         f: impl FnOnce() -> Result<Option<Account>, E>,
@@ -645,12 +625,7 @@ impl ExecutionCache {
     }
 
     /// Insert storage value into cache.
-    pub(crate) fn insert_storage(
-        &self,
-        address: Address,
-        key: StorageKey,
-        value: Option<StorageValue>,
-    ) {
+    pub fn insert_storage(&self, address: Address, key: StorageKey, value: Option<StorageValue>) {
         self.storage_cache.insert((address, key), value.unwrap_or_default());
     }
 
@@ -683,7 +658,8 @@ impl ExecutionCache {
     ///
     /// Returns an error if the state updates are inconsistent and should be discarded.
     #[instrument(level = "debug", target = "engine::caching", skip_all)]
-    pub(crate) fn insert_state(&self, state_updates: &BundleState) -> Result<(), ()> {
+    #[expect(clippy::result_unit_err)]
+    pub fn insert_state(&self, state_updates: &BundleState) -> Result<(), ()> {
         let _enter =
             debug_span!(target: "engine::tree", "contracts", len = state_updates.contracts.len())
                 .entered();
@@ -792,7 +768,7 @@ impl ExecutionCache {
 /// A saved cache that has been used for executing a specific block, which has been updated for its
 /// execution.
 #[derive(Debug, Clone)]
-pub(crate) struct SavedCache {
+pub struct SavedCache {
     /// The hash of the block these caches were used to execute.
     hash: B256,
 
@@ -812,43 +788,43 @@ pub(crate) struct SavedCache {
 
 impl SavedCache {
     /// Creates a new instance with the internals
-    pub(super) fn new(hash: B256, caches: ExecutionCache, metrics: CachedStateMetrics) -> Self {
+    pub fn new(hash: B256, caches: ExecutionCache, metrics: CachedStateMetrics) -> Self {
         Self { hash, caches, metrics, usage_guard: Arc::new(()), disable_cache_metrics: false }
     }
 
     /// Sets whether to disable cache metrics recording.
-    pub(super) const fn with_disable_cache_metrics(mut self, disable: bool) -> Self {
+    pub const fn with_disable_cache_metrics(mut self, disable: bool) -> Self {
         self.disable_cache_metrics = disable;
         self
     }
 
     /// Returns the hash for this cache
-    pub(crate) const fn executed_block_hash(&self) -> B256 {
+    pub const fn executed_block_hash(&self) -> B256 {
         self.hash
     }
 
     /// Splits the cache into its caches, metrics, and `disable_cache_metrics` flag, consuming it.
-    pub(crate) fn split(self) -> (ExecutionCache, CachedStateMetrics, bool) {
+    pub fn split(self) -> (ExecutionCache, CachedStateMetrics, bool) {
         (self.caches, self.metrics, self.disable_cache_metrics)
     }
 
     /// Returns true if the cache is available for use (no other tasks are currently using it).
-    pub(crate) fn is_available(&self) -> bool {
+    pub fn is_available(&self) -> bool {
         Arc::strong_count(&self.usage_guard) == 1
     }
 
     /// Returns the current strong count of the usage guard.
-    pub(crate) fn usage_count(&self) -> usize {
+    pub fn usage_count(&self) -> usize {
         Arc::strong_count(&self.usage_guard)
     }
 
     /// Returns the [`ExecutionCache`] belonging to the tracked hash.
-    pub(crate) const fn cache(&self) -> &ExecutionCache {
+    pub const fn cache(&self) -> &ExecutionCache {
         &self.caches
     }
 
     /// Returns the metrics associated with this cache.
-    pub(crate) const fn metrics(&self) -> &CachedStateMetrics {
+    pub const fn metrics(&self) -> &CachedStateMetrics {
         &self.metrics
     }
 

crates/engine/tree/src/tree/metrics.rs

@@ -13,13 +13,13 @@ use std::time::{Duration, Instant};
 
 /// Metrics for the `EngineApi`.
 #[derive(Debug, Default)]
-pub(crate) struct EngineApiMetrics {
+pub struct EngineApiMetrics {
     /// Engine API-specific metrics.
-    pub(crate) engine: EngineMetrics,
+    pub engine: EngineMetrics,
     /// Block executor metrics.
-    pub(crate) executor: ExecutorMetrics,
+    pub executor: ExecutorMetrics,
     /// Metrics for block validation
-    pub(crate) block_validation: BlockValidationMetrics,
+    pub block_validation: BlockValidationMetrics,
     /// Canonical chain and reorg related metrics
     pub tree: TreeMetrics,
     /// Metrics for EIP-7928 Block-Level Access Lists (BAL).
@@ -32,7 +32,7 @@ impl EngineApiMetrics {
     ///
     /// This method updates metrics for execution time, gas usage, and the number
     /// of accounts, storage slots and bytecodes updated.
-    pub(crate) fn record_block_execution<R>(
+    pub fn record_block_execution<R>(
         &self,
         output: &BlockExecutionOutput<R>,
         execution_duration: Duration,
@@ -59,27 +59,27 @@ impl EngineApiMetrics {
     }
 
     /// Returns a reference to the executor metrics for use in state hooks.
-    pub(crate) const fn executor_metrics(&self) -> &ExecutorMetrics {
+    pub const fn executor_metrics(&self) -> &ExecutorMetrics {
         &self.executor
     }
 
     /// Records the duration of block pre-execution changes (e.g., beacon root update).
-    pub(crate) fn record_pre_execution(&self, elapsed: Duration) {
+    pub fn record_pre_execution(&self, elapsed: Duration) {
         self.executor.pre_execution_histogram.record(elapsed);
     }
 
     /// Records the duration of block post-execution changes (e.g., finalization).
-    pub(crate) fn record_post_execution(&self, elapsed: Duration) {
+    pub fn record_post_execution(&self, elapsed: Duration) {
         self.executor.post_execution_histogram.record(elapsed);
     }
 
     /// Records the time spent waiting for the next transaction from the iterator.
-    pub(crate) fn record_transaction_wait(&self, elapsed: Duration) {
+    pub fn record_transaction_wait(&self, elapsed: Duration) {
         self.executor.transaction_wait_histogram.record(elapsed);
     }
 
     /// Records the duration of a single transaction execution.
-    pub(crate) fn record_transaction_execution(&self, elapsed: Duration) {
+    pub fn record_transaction_execution(&self, elapsed: Duration) {
         self.executor.transaction_execution_histogram.record(elapsed);
     }
 }
@@ -87,7 +87,7 @@ impl EngineApiMetrics {
 /// Metrics for the entire blockchain tree
 #[derive(Metrics)]
 #[metrics(scope = "blockchain_tree")]
-pub(crate) struct TreeMetrics {
+pub struct TreeMetrics {
     /// The highest block number in the canonical chain
     pub canonical_chain_height: Gauge,
     /// The number of reorgs
@@ -103,7 +103,7 @@ pub(crate) struct TreeMetrics {
 /// Metrics for the `EngineApi`.
 #[derive(Metrics)]
 #[metrics(scope = "consensus.engine.beacon")]
-pub(crate) struct EngineMetrics {
+pub struct EngineMetrics {
     /// Engine API forkchoiceUpdated response type metrics
     #[metric(skip)]
     pub(crate) forkchoice_updated: ForkchoiceUpdatedMetrics,
@@ -336,42 +336,42 @@ pub(crate) struct BalMetrics {
 /// Metrics for non-execution related block validation.
 #[derive(Metrics, Clone)]
 #[metrics(scope = "sync.block_validation")]
-pub(crate) struct BlockValidationMetrics {
+pub struct BlockValidationMetrics {
     /// Total number of storage tries updated in the state root calculation
-    pub(crate) state_root_storage_tries_updated_total: Counter,
+    pub state_root_storage_tries_updated_total: Counter,
     /// Total number of times the parallel state root computation fell back to regular.
-    pub(crate) state_root_parallel_fallback_total: Counter,
+    pub state_root_parallel_fallback_total: Counter,
     /// Total number of times the state root task failed but the fallback succeeded.
-    pub(crate) state_root_task_fallback_success_total: Counter,
+    pub state_root_task_fallback_success_total: Counter,
     /// Latest state root duration, ie the time spent blocked waiting for the state root.
-    pub(crate) state_root_duration: Gauge,
+    pub state_root_duration: Gauge,
     /// Histogram for state root duration ie the time spent blocked waiting for the state root
-    pub(crate) state_root_histogram: Histogram,
+    pub state_root_histogram: Histogram,
     /// Histogram of deferred trie computation duration.
-    pub(crate) deferred_trie_compute_duration: Histogram,
+    pub deferred_trie_compute_duration: Histogram,
     /// Payload conversion and validation latency
-    pub(crate) payload_validation_duration: Gauge,
+    pub payload_validation_duration: Gauge,
     /// Histogram of payload validation latency
-    pub(crate) payload_validation_histogram: Histogram,
+    pub payload_validation_histogram: Histogram,
     /// Payload processor spawning duration
-    pub(crate) spawn_payload_processor: Histogram,
+    pub spawn_payload_processor: Histogram,
     /// Post-execution validation duration
-    pub(crate) post_execution_validation_duration: Histogram,
+    pub post_execution_validation_duration: Histogram,
     /// Total duration of the new payload call
-    pub(crate) total_duration: Histogram,
+    pub total_duration: Histogram,
     /// Size of `HashedPostStateSorted` (`total_len`)
-    pub(crate) hashed_post_state_size: Histogram,
+    pub hashed_post_state_size: Histogram,
     /// Size of `TrieUpdatesSorted` (`total_len`)
-    pub(crate) trie_updates_sorted_size: Histogram,
+    pub trie_updates_sorted_size: Histogram,
     /// Size of `AnchoredTrieInput` overlay `TrieUpdatesSorted` (`total_len`)
-    pub(crate) anchored_overlay_trie_updates_size: Histogram,
+    pub anchored_overlay_trie_updates_size: Histogram,
     /// Size of `AnchoredTrieInput` overlay `HashedPostStateSorted` (`total_len`)
-    pub(crate) anchored_overlay_hashed_state_size: Histogram,
+    pub anchored_overlay_hashed_state_size: Histogram,
 }
 
 impl BlockValidationMetrics {
     /// Records a new state root time, updating both the histogram and state root gauge
-    pub(crate) fn record_state_root(&self, trie_output: &TrieUpdates, elapsed_as_secs: f64) {
+    pub fn record_state_root(&self, trie_output: &TrieUpdates, elapsed_as_secs: f64) {
         self.state_root_storage_tries_updated_total
             .increment(trie_output.storage_tries_ref().len() as u64);
         self.state_root_duration.set(elapsed_as_secs);
@@ -380,7 +380,7 @@ impl BlockValidationMetrics {
 
     /// Records a new payload validation time, updating both the histogram and the payload
     /// validation gauge
-    pub(crate) fn record_payload_validation(&self, elapsed_as_secs: f64) {
+    pub fn record_payload_validation(&self, elapsed_as_secs: f64) {
         self.payload_validation_duration.set(elapsed_as_secs);
         self.payload_validation_histogram.record(elapsed_as_secs);
     }

crates/engine/tree/src/tree/mod.rs

@@ -3,7 +3,7 @@ use crate::{
     chain::FromOrchestrator,
     engine::{DownloadRequest, EngineApiEvent, EngineApiKind, EngineApiRequest, FromEngine},
     persistence::PersistenceHandle,
-    tree::{error::InsertPayloadError, metrics::EngineApiMetrics, payload_validator::TreeCtx},
+    tree::{error::InsertPayloadError, payload_validator::TreeCtx},
 };
 use alloy_consensus::BlockHeader;
 use alloy_eips::{eip1898::BlockWithParent, merge::EPOCH_SLOTS, BlockNumHash, NumHash};
@@ -30,7 +30,7 @@ use reth_payload_primitives::{
 };
 use reth_primitives_traits::{NodePrimitives, RecoveredBlock, SealedBlock, SealedHeader};
 use reth_provider::{
-    BlockExecutionOutput, BlockExecutionResult, BlockNumReader, BlockReader, ChangeSetReader,
+    BlockExecutionOutput, BlockExecutionResult, BlockReader, ChangeSetReader,
     DatabaseProviderFactory, HashedPostStateProvider, ProviderError, StageCheckpointReader,
     StateProviderBox, StateProviderFactory, StateReader, StorageChangeSetReader,
     TransactionVariant,
@@ -55,18 +55,19 @@ pub mod error;
 pub mod instrumented_state;
 mod invalid_headers;
 mod metrics;
-mod payload_processor;
+pub mod payload_processor;
 pub mod payload_validator;
 mod persistence_state;
 pub mod precompile_cache;
 #[cfg(test)]
 mod tests;
-#[expect(unused)]
 mod trie_updates;
 
 use crate::tree::error::AdvancePersistenceError;
 pub use block_buffer::BlockBuffer;
+pub use cached_state::{CachedStateMetrics, CachedStateProvider, ExecutionCache, SavedCache};
 pub use invalid_headers::InvalidHeaderCache;
+pub use metrics::EngineApiMetrics;
 pub use payload_processor::*;
 pub use payload_validator::{BasicEngineValidator, EngineValidator};
 pub use persistence_state::PersistenceState;
@@ -158,6 +159,16 @@ impl<N: NodePrimitives> EngineApiTreeState<N> {
             forkchoice_state_tracker: ForkchoiceStateTracker::default(),
         }
     }
+
+    /// Returns a reference to the tree state.
+    pub const fn tree_state(&self) -> &TreeState<N> {
+        &self.tree_state
+    }
+
+    /// Returns true if the block has been marked as invalid.
+    pub fn has_invalid_header(&mut self, hash: &B256) -> bool {
+        self.invalid_headers.get(hash).is_some()
+    }
 }
 
 /// The outcome of a tree operation.
@@ -321,11 +332,10 @@ where
         + HashedPostStateProvider
         + Clone
         + 'static,
-    <P as DatabaseProviderFactory>::Provider: BlockReader<Block = N::Block, Header = N::BlockHeader>
+    P::Provider: BlockReader<Block = N::Block, Header = N::BlockHeader>
         + StageCheckpointReader
         + ChangeSetReader
-        + StorageChangeSetReader
-        + BlockNumReader,
+        + StorageChangeSetReader,
     C: ConfigureEvm<Primitives = N> + 'static,
     T: PayloadTypes<BuiltPayload: BuiltPayload<Primitives = N>>,
     V: EngineValidator<T>,
@@ -1406,7 +1416,20 @@ where
         );
         self.changeset_cache.evict(eviction_threshold);
 
+        // Invalidate cached overlay since the anchor has changed
+        self.state.tree_state.invalidate_cached_overlay();
+
         self.on_new_persisted_block()?;
+
+        // Re-prepare overlay for the current canonical head with the new anchor.
+        // Spawn a background task to trigger computation so it's ready when the next payload
+        // arrives.
+        if let Some(overlay) = self.state.tree_state.prepare_canonical_overlay() {
+            rayon::spawn(move || {
+                let _ = overlay.get();
+            });
+        }
+
         Ok(())
     }
 
@@ -2589,19 +2612,27 @@ where
         let block_num_hash = block_id.block;
         debug!(target: "engine::tree", block=?block_num_hash, parent = ?block_id.parent, "Inserting new block into tree");
 
-        match self.sealed_header_by_hash(block_num_hash.hash) {
-            Err(err) => {
-                let block = convert_to_block(self, input)?;
-                return Err(InsertBlockError::new(block, err.into()).into());
+        // Check if block already exists - first in memory, then DB only if it could be persisted
+        if self.state.tree_state.sealed_header_by_hash(&block_num_hash.hash).is_some() {
+            convert_to_block(self, input)?;
+            return Ok(InsertPayloadOk::AlreadySeen(BlockStatus::Valid));
+        }
+
+        // Only query DB if block could be persisted (number <= last persisted block).
+        // New blocks from CL always have number > last persisted, so skip DB lookup for them.
+        if block_num_hash.number <= self.persistence_state.last_persisted_block.number {
+            match self.provider.sealed_header_by_hash(block_num_hash.hash) {
+                Err(err) => {
+                    let block = convert_to_block(self, input)?;
+                    return Err(InsertBlockError::new(block, err.into()).into());
+                }
+                Ok(Some(_)) => {
+                    convert_to_block(self, input)?;
+                    return Ok(InsertPayloadOk::AlreadySeen(BlockStatus::Valid));
+                }
+                Ok(None) => {}
             }
-            Ok(Some(_)) => {
-                // We now assume that we already have this block in the tree. However, we need to
-                // run the conversion to ensure that the block hash is valid.
-                convert_to_block(self, input)?;
-                return Ok(InsertPayloadOk::AlreadySeen(BlockStatus::Valid))
-            }
-            _ => {}
-        };
+        }
 
         // Ensure that the parent state is available.
         match self.state_provider_builder(block_id.parent) {

crates/engine/tree/src/tree/payload_processor/bal.rs

@@ -21,7 +21,7 @@ pub fn total_slots(bal: &BlockAccessList) -> usize {
 /// first, followed by read-only slots. The iterator intelligently skips accounts and slots
 /// outside the specified range for efficient traversal.
 #[derive(Debug)]
-pub(crate) struct BALSlotIter<'a> {
+pub struct BALSlotIter<'a> {
     bal: &'a BlockAccessList,
     range: Range<usize>,
     current_index: usize,
@@ -34,7 +34,7 @@ pub(crate) struct BALSlotIter<'a> {
 
 impl<'a> BALSlotIter<'a> {
     /// Creates a new iterator over storage slots within the specified range.
-    pub(crate) fn new(bal: &'a BlockAccessList, range: Range<usize>) -> Self {
+    pub fn new(bal: &'a BlockAccessList, range: Range<usize>) -> Self {
         let mut iter = Self { bal, range, current_index: 0, account_idx: 0, slot_idx: 0 };
         iter.skip_to_range_start();
         iter

crates/engine/tree/src/tree/payload_processor/mod.rs

@@ -7,14 +7,14 @@ use crate::tree::{
         prewarm::{PrewarmCacheTask, PrewarmContext, PrewarmMode, PrewarmTaskEvent},
         sparse_trie::StateRootComputeOutcome,
     },
-    sparse_trie::SparseTrieTask,
+    sparse_trie::{SparseTrieCacheTask, SparseTrieTask, SpawnedSparseTrieTask},
     StateProviderBuilder, TreeConfig,
 };
 use alloy_eip7928::BlockAccessList;
 use alloy_eips::eip1898::BlockWithParent;
 use alloy_evm::block::StateChangeSource;
 use alloy_primitives::B256;
-use crossbeam_channel::Sender as CrossbeamSender;
+use crossbeam_channel::{Receiver as CrossbeamReceiver, Sender as CrossbeamSender};
 use executor::WorkloadExecutor;
 use metrics::Counter;
 use multiproof::{SparseTrieUpdate, *};
@@ -39,10 +39,7 @@ use reth_trie_parallel::{
     proof_task::{ProofTaskCtx, ProofWorkerHandle},
     root::ParallelStateRootError,
 };
-use reth_trie_sparse::{
-    provider::{TrieNodeProvider, TrieNodeProviderFactory},
-    ClearedSparseStateTrie, RevealableSparseTrie, SparseStateTrie,
-};
+use reth_trie_sparse::{RevealableSparseTrie, SparseStateTrie};
 use reth_trie_sparse_parallel::{ParallelSparseTrie, ParallelismThresholds};
 use std::{
     collections::BTreeMap,
@@ -59,10 +56,13 @@ use tracing::{debug, debug_span, instrument, warn, Span};
 pub mod bal;
 pub mod executor;
 pub mod multiproof;
+mod preserved_sparse_trie;
 pub mod prewarm;
 pub mod receipt_root_task;
 pub mod sparse_trie;
 
+use preserved_sparse_trie::{PreservedSparseTrie, SharedPreservedSparseTrie};
+
 /// Default parallelism thresholds to use with the [`ParallelSparseTrie`].
 ///
 /// These values were determined by performing benchmarks using gradually increasing values to judge
@@ -125,13 +125,16 @@ where
     precompile_cache_disabled: bool,
     /// Precompile cache map.
     precompile_cache_map: PrecompileCacheMap<SpecFor<Evm>>,
-    /// A cleared `SparseStateTrie`, kept around to be reused for the state root computation so
-    /// that allocations can be minimized.
-    sparse_state_trie: Arc<
-        parking_lot::Mutex<Option<ClearedSparseStateTrie<ParallelSparseTrie, ParallelSparseTrie>>>,
-    >,
+    /// A pruned `SparseStateTrie`, kept around as a cache of already revealed trie nodes and to
+    /// re-use allocated memory. Stored with the block hash it was computed for to enable trie
+    /// preservation across sequential payload validations.
+    sparse_state_trie: SharedPreservedSparseTrie,
     /// Maximum concurrency for prewarm task.
     prewarm_max_concurrency: usize,
+    /// Sparse trie prune depth.
+    sparse_trie_prune_depth: usize,
+    /// Maximum storage tries to retain after pruning.
+    sparse_trie_max_storage_tries: usize,
     /// Whether to disable cache metrics recording.
     disable_cache_metrics: bool,
 }
@@ -142,7 +145,7 @@ where
     Evm: ConfigureEvm<Primitives = N>,
 {
     /// Returns a reference to the workload executor driving payload tasks.
-    pub(super) const fn executor(&self) -> &WorkloadExecutor {
+    pub const fn executor(&self) -> &WorkloadExecutor {
         &self.executor
     }
 
@@ -163,8 +166,10 @@ where
             disable_state_cache: config.disable_state_cache(),
             precompile_cache_disabled: config.precompile_cache_disabled(),
             precompile_cache_map,
-            sparse_state_trie: Arc::default(),
+            sparse_state_trie: SharedPreservedSparseTrie::default(),
             prewarm_max_concurrency: config.prewarm_max_concurrency(),
+            sparse_trie_prune_depth: config.sparse_trie_prune_depth(),
+            sparse_trie_max_storage_tries: config.sparse_trie_max_storage_tries(),
             disable_cache_metrics: config.disable_cache_metrics(),
         }
     }
@@ -230,15 +235,17 @@ where
             + 'static,
     {
         // start preparing transactions immediately
-        let (prewarm_rx, execution_rx, transaction_count_hint) =
-            self.spawn_tx_iterator(transactions);
+        let (prewarm_rx, execution_rx) = self.spawn_tx_iterator(transactions);
 
         let span = Span::current();
         let (to_sparse_trie, sparse_trie_rx) = channel();
         let (to_multi_proof, from_multi_proof) = crossbeam_channel::unbounded();
 
         // Extract V2 proofs flag early so we can pass it to prewarm
-        let v2_proofs_enabled = config.enable_proof_v2();
+        let v2_proofs_enabled = !config.disable_proof_v2();
+
+        // Capture parent_state_root before env is moved into spawn_caching_with
+        let parent_state_root = env.parent_state_root;
 
         // Handle BAL-based optimization if available
         let prewarm_handle = if let Some(bal) = bal {
@@ -252,7 +259,6 @@ where
             self.spawn_caching_with(
                 env,
                 prewarm_rx,
-                transaction_count_hint,
                 provider_builder.clone(),
                 None, // Don't send proof targets when BAL is present
                 Some(bal),
@@ -263,7 +269,6 @@ where
             self.spawn_caching_with(
                 env,
                 prewarm_rx,
-                transaction_count_hint,
                 provider_builder.clone(),
                 Some(to_multi_proof.clone()),
                 None,
@@ -283,39 +288,46 @@ where
             v2_proofs_enabled,
         );
 
-        let multi_proof_task = MultiProofTask::new(
-            proof_handle.clone(),
-            to_sparse_trie,
-            config
-                .multiproof_chunking_enabled()
-                .then_some(config.effective_multiproof_chunk_size()),
-            to_multi_proof.clone(),
-            from_multi_proof,
-        )
-        .with_v2_proofs_enabled(v2_proofs_enabled);
+        if !config.enable_sparse_trie_as_cache() {
+            let multi_proof_task = MultiProofTask::new(
+                proof_handle.clone(),
+                to_sparse_trie,
+                config.multiproof_chunking_enabled().then_some(config.multiproof_chunk_size()),
+                to_multi_proof.clone(),
+                from_multi_proof.clone(),
+            )
+            .with_v2_proofs_enabled(v2_proofs_enabled);
 
-        // spawn multi-proof task
-        let parent_span = span.clone();
-        let saved_cache = prewarm_handle.saved_cache.clone();
-        self.executor.spawn_blocking(move || {
-            let _enter = parent_span.entered();
-            // Build a state provider for the multiproof task
-            let provider = provider_builder.build().expect("failed to build provider");
-            let provider = if let Some(saved_cache) = saved_cache {
-                let (cache, metrics, _disable_metrics) = saved_cache.split();
-                Box::new(CachedStateProvider::new(provider, cache, metrics))
-                    as Box<dyn StateProvider>
-            } else {
-                Box::new(provider)
-            };
-            multi_proof_task.run(provider);
-        });
+            // spawn multi-proof task
+            let parent_span = span.clone();
+            let saved_cache = prewarm_handle.saved_cache.clone();
+            self.executor.spawn_blocking(move || {
+                let _enter = parent_span.entered();
+                // Build a state provider for the multiproof task
+                let provider = provider_builder.build().expect("failed to build provider");
+                let provider = if let Some(saved_cache) = saved_cache {
+                    let (cache, metrics, _disable_metrics) = saved_cache.split();
+                    Box::new(CachedStateProvider::new(provider, cache, metrics))
+                        as Box<dyn StateProvider>
+                } else {
+                    Box::new(provider)
+                };
+                multi_proof_task.run(provider);
+            });
+        }
 
         // wire the sparse trie to the state root response receiver
         let (state_root_tx, state_root_rx) = channel();
 
         // Spawn the sparse trie task using any stored trie and parallel trie configuration.
-        self.spawn_sparse_trie_task(sparse_trie_rx, proof_handle, state_root_tx);
+        self.spawn_sparse_trie_task(
+            sparse_trie_rx,
+            proof_handle,
+            state_root_tx,
+            from_multi_proof,
+            config,
+            parent_state_root,
+        );
 
         PayloadHandle {
             to_multi_proof: Some(to_multi_proof),
@@ -330,7 +342,7 @@ where
     ///
     /// Returns a [`PayloadHandle`] to communicate with the task.
     #[instrument(level = "debug", target = "engine::tree::payload_processor", skip_all)]
-    pub(super) fn spawn_cache_exclusive<P, I: ExecutableTxIterator<Evm>>(
+    pub fn spawn_cache_exclusive<P, I: ExecutableTxIterator<Evm>>(
         &self,
         env: ExecutionEnv<Evm>,
         transactions: I,
@@ -340,10 +352,10 @@ where
     where
         P: BlockReader + StateProviderFactory + StateReader + Clone + 'static,
     {
-        let (prewarm_rx, execution_rx, size_hint) = self.spawn_tx_iterator(transactions);
+        let (prewarm_rx, execution_rx) = self.spawn_tx_iterator(transactions);
         // This path doesn't use multiproof, so V2 proofs flag doesn't matter
         let prewarm_handle =
-            self.spawn_caching_with(env, prewarm_rx, size_hint, provider_builder, None, bal, false);
+            self.spawn_caching_with(env, prewarm_rx, provider_builder, None, bal, false);
         PayloadHandle {
             to_multi_proof: None,
             prewarm_handle,
@@ -361,19 +373,15 @@ where
     ) -> (
         mpsc::Receiver<WithTxEnv<TxEnvFor<Evm>, I::Recovered>>,
         mpsc::Receiver<Result<WithTxEnv<TxEnvFor<Evm>, I::Recovered>, I::Error>>,
-        usize,
     ) {
-        let (transactions, convert) = transactions.into();
-        let transactions = transactions.into_par_iter();
-        let transaction_count_hint = transactions.len();
-
         let (ooo_tx, ooo_rx) = mpsc::channel();
         let (prewarm_tx, prewarm_rx) = mpsc::channel();
         let (execute_tx, execute_rx) = mpsc::channel();
 
         // Spawn a task that `convert`s all transactions in parallel and sends them out-of-order.
-        self.executor.spawn_blocking(move || {
-            transactions.enumerate().for_each_with(ooo_tx, |ooo_tx, (idx, tx)| {
+        rayon::spawn(move || {
+            let (transactions, convert) = transactions.into();
+            transactions.into_par_iter().enumerate().for_each_with(ooo_tx, |ooo_tx, (idx, tx)| {
                 let tx = convert(tx);
                 let tx = tx.map(|tx| {
                     let (tx_env, tx) = tx.into_parts();
@@ -409,16 +417,14 @@ where
             }
         });
 
-        (prewarm_rx, execute_rx, transaction_count_hint)
+        (prewarm_rx, execute_rx)
     }
 
     /// Spawn prewarming optionally wired to the multiproof task for target updates.
-    #[expect(clippy::too_many_arguments)]
     fn spawn_caching_with<P>(
         &self,
         env: ExecutionEnv<Evm>,
         mut transactions: mpsc::Receiver<impl ExecutableTxFor<Evm> + Clone + Send + 'static>,
-        transaction_count_hint: usize,
         provider_builder: StateProviderBuilder<N, P>,
         to_multi_proof: Option<CrossbeamSender<MultiProofMessage>>,
         bal: Option<Arc<BlockAccessList>>,
@@ -453,7 +459,6 @@ where
             self.execution_cache.clone(),
             prewarm_ctx,
             to_multi_proof,
-            transaction_count_hint,
             self.prewarm_max_concurrency,
         );
 
@@ -494,64 +499,121 @@ where
     }
 
     /// Spawns the [`SparseTrieTask`] for this payload processor.
+    ///
+    /// The trie is preserved when the new payload is a child of the previous one.
     #[instrument(level = "debug", target = "engine::tree::payload_processor", skip_all)]
-    fn spawn_sparse_trie_task<BPF>(
+    fn spawn_sparse_trie_task(
         &self,
         sparse_trie_rx: mpsc::Receiver<SparseTrieUpdate>,
-        proof_worker_handle: BPF,
+        proof_worker_handle: ProofWorkerHandle,
         state_root_tx: mpsc::Sender<Result<StateRootComputeOutcome, ParallelStateRootError>>,
-    ) where
-        BPF: TrieNodeProviderFactory + Clone + Send + Sync + 'static,
-        BPF::AccountNodeProvider: TrieNodeProvider + Send + Sync,
-        BPF::StorageNodeProvider: TrieNodeProvider + Send + Sync,
-    {
-        let cleared_sparse_trie = Arc::clone(&self.sparse_state_trie);
+        from_multi_proof: CrossbeamReceiver<MultiProofMessage>,
+        config: &TreeConfig,
+        parent_state_root: B256,
+    ) {
+        let preserved_sparse_trie = self.sparse_state_trie.clone();
         let trie_metrics = self.trie_metrics.clone();
         let span = Span::current();
+        let disable_sparse_trie_as_cache = !config.enable_sparse_trie_as_cache();
+        let prune_depth = self.sparse_trie_prune_depth;
+        let max_storage_tries = self.sparse_trie_max_storage_tries;
 
         self.executor.spawn_blocking(move || {
             let _enter = span.entered();
 
-            // Reuse a stored SparseStateTrie, or create a new one using the desired configuration
-            // if there's none to reuse.
-            let sparse_state_trie = cleared_sparse_trie.lock().take().unwrap_or_else(|| {
-                let default_trie = RevealableSparseTrie::blind_from(
-                    ParallelSparseTrie::default()
-                        .with_parallelism_thresholds(PARALLEL_SPARSE_TRIE_PARALLELISM_THRESHOLDS),
-                );
-                ClearedSparseStateTrie::from_state_trie(
+            // Reuse a stored SparseStateTrie if available, applying continuation logic.
+            // If this payload's parent state root matches the preserved trie's anchor,
+            // we can reuse the pruned trie structure. Otherwise, we clear the trie but
+            // keep allocations.
+            let sparse_state_trie = preserved_sparse_trie
+                .take()
+                .map(|preserved| preserved.into_trie_for(parent_state_root))
+                .unwrap_or_else(|| {
+                    debug!(
+                        target: "engine::tree::payload_processor",
+                        "Creating new sparse trie - no preserved trie available"
+                    );
+                    let default_trie = RevealableSparseTrie::blind_from(
+                        ParallelSparseTrie::default().with_parallelism_thresholds(
+                            PARALLEL_SPARSE_TRIE_PARALLELISM_THRESHOLDS,
+                        ),
+                    );
                     SparseStateTrie::new()
                         .with_accounts_trie(default_trie.clone())
                         .with_default_storage_trie(default_trie)
-                        .with_updates(true),
-                )
-            });
+                        .with_updates(true)
+                });
 
-            let task =
-                SparseTrieTask::<_, ParallelSparseTrie, ParallelSparseTrie>::new_with_cleared_trie(
+            let mut task = if disable_sparse_trie_as_cache {
+                SpawnedSparseTrieTask::Cleared(SparseTrieTask::new(
                     sparse_trie_rx,
                     proof_worker_handle,
-                    trie_metrics,
+                    trie_metrics.clone(),
                     sparse_state_trie,
+                ))
+            } else {
+                SpawnedSparseTrieTask::Cached(SparseTrieCacheTask::new_with_trie(
+                    from_multi_proof,
+                    proof_worker_handle,
+                    trie_metrics.clone(),
+                    sparse_state_trie,
+                ))
+            };
+
+            let result = task.run();
+            // Capture the computed state_root before sending the result
+            let computed_state_root = result.as_ref().ok().map(|outcome| outcome.state_root);
+
+            // Acquire the guard before sending the result to prevent a race condition:
+            // Without this, the next block could start after send() but before store(),
+            // causing take() to return None and forcing it to create a new empty trie
+            // instead of reusing the preserved one. Holding the guard ensures the next
+            // block's take() blocks until we've stored the trie for reuse.
+            let mut guard = preserved_sparse_trie.lock();
+
+            // Send state root computation result - next block may start but will block on take()
+            if state_root_tx.send(result).is_err() {
+                // Receiver dropped - payload was likely invalid or cancelled.
+                // Clear the trie instead of preserving potentially invalid state.
+                debug!(
+                    target: "engine::tree::payload_processor",
+                    "State root receiver dropped, clearing trie"
                 );
+                let trie = task.into_cleared_trie(
+                    SPARSE_TRIE_MAX_NODES_SHRINK_CAPACITY,
+                    SPARSE_TRIE_MAX_VALUES_SHRINK_CAPACITY,
+                );
+                guard.store(PreservedSparseTrie::cleared(trie));
+                return;
+            }
 
-            let (result, trie) = task.run();
-            // Send state root computation result
-            let _ = state_root_tx.send(result);
-
-            // Clear the SparseStateTrie, shrink, and replace it back into the mutex _after_ sending
-            // results to the next step, so that time spent clearing doesn't block the step after
-            // this one.
-            let _enter = debug_span!(target: "engine::tree::payload_processor", "clear").entered();
-            let mut cleared_trie = ClearedSparseStateTrie::from_state_trie(trie);
-
-            // Shrink the sparse trie so that we don't have ever increasing memory.
-            cleared_trie.shrink_to(
-                SPARSE_TRIE_MAX_NODES_SHRINK_CAPACITY,
-                SPARSE_TRIE_MAX_VALUES_SHRINK_CAPACITY,
-            );
-
-            cleared_sparse_trie.lock().replace(cleared_trie);
+            // Only preserve the trie as anchored if computation succeeded.
+            // A failed computation may have left the trie in a partially updated state.
+            let _enter =
+                debug_span!(target: "engine::tree::payload_processor", "preserve").entered();
+            if let Some(state_root) = computed_state_root {
+                let start = std::time::Instant::now();
+                let trie = task.into_trie_for_reuse(
+                    prune_depth,
+                    max_storage_tries,
+                    SPARSE_TRIE_MAX_NODES_SHRINK_CAPACITY,
+                    SPARSE_TRIE_MAX_VALUES_SHRINK_CAPACITY,
+                );
+                trie_metrics
+                    .into_trie_for_reuse_duration_histogram
+                    .record(start.elapsed().as_secs_f64());
+                guard.store(PreservedSparseTrie::anchored(trie, state_root));
+            } else {
+                debug!(
+                    target: "engine::tree::payload_processor",
+                    "State root computation failed, clearing trie"
+                );
+                let trie = task.into_cleared_trie(
+                    SPARSE_TRIE_MAX_NODES_SHRINK_CAPACITY,
+                    SPARSE_TRIE_MAX_VALUES_SHRINK_CAPACITY,
+                );
+                guard.store(PreservedSparseTrie::cleared(trie));
+            }
         });
     }
 
@@ -562,7 +624,7 @@ where
     ///
     /// The cache enables subsequent blocks to reuse account, storage, and bytecode data without
     /// hitting the database, maintaining performance consistency.
-    pub(crate) fn on_inserted_executed_block(
+    pub fn on_inserted_executed_block(
         &self,
         block_with_parent: BlockWithParent,
         bundle_state: &BundleState,
@@ -659,19 +721,19 @@ impl<Tx, Err, R: Send + Sync + 'static> PayloadHandle<Tx, Err, R> {
     }
 
     /// Returns a clone of the caches used by prewarming
-    pub(super) fn caches(&self) -> Option<ExecutionCache> {
+    pub fn caches(&self) -> Option<ExecutionCache> {
         self.prewarm_handle.saved_cache.as_ref().map(|cache| cache.cache().clone())
     }
 
     /// Returns a clone of the cache metrics used by prewarming
-    pub(super) fn cache_metrics(&self) -> Option<CachedStateMetrics> {
+    pub fn cache_metrics(&self) -> Option<CachedStateMetrics> {
         self.prewarm_handle.saved_cache.as_ref().map(|cache| cache.metrics().clone())
     }
 
     /// Terminates the pre-warming transaction processing.
     ///
     /// Note: This does not terminate the task yet.
-    pub(super) fn stop_prewarming_execution(&self) {
+    pub fn stop_prewarming_execution(&self) {
         self.prewarm_handle.stop_prewarming_execution()
     }
 
@@ -682,7 +744,7 @@ impl<Tx, Err, R: Send + Sync + 'static> PayloadHandle<Tx, Err, R> {
     /// path without cloning the expensive `BundleState`.
     ///
     /// Returns a sender for the channel that should be notified on block validation success.
-    pub(super) fn terminate_caching(
+    pub fn terminate_caching(
         &mut self,
         execution_outcome: Option<Arc<BlockExecutionOutput<R>>>,
     ) -> Option<mpsc::Sender<()>> {
@@ -702,7 +764,7 @@ impl<Tx, Err, R: Send + Sync + 'static> PayloadHandle<Tx, Err, R> {
 /// Generic over `R` (receipt type) to allow sharing `Arc<ExecutionOutcome<R>>` with the
 /// prewarm task without cloning the expensive `BundleState`.
 #[derive(Debug)]
-pub(crate) struct CacheTaskHandle<R> {
+pub struct CacheTaskHandle<R> {
     /// The shared cache the task operates with.
     saved_cache: Option<SavedCache>,
     /// Channel to the spawned prewarm task if any
@@ -713,7 +775,7 @@ impl<R: Send + Sync + 'static> CacheTaskHandle<R> {
     /// Terminates the pre-warming transaction processing.
     ///
     /// Note: This does not terminate the task yet.
-    pub(super) fn stop_prewarming_execution(&self) {
+    pub fn stop_prewarming_execution(&self) {
         self.to_prewarm_task
             .as_ref()
             .map(|tx| tx.send(PrewarmTaskEvent::TerminateTransactionExecution).ok());
@@ -724,7 +786,7 @@ impl<R: Send + Sync + 'static> CacheTaskHandle<R> {
     /// If the [`BlockExecutionOutput`] is provided it will update the shared cache using its
     /// bundle state. Using `Arc<ExecutionOutcome>` avoids cloning the expensive `BundleState`.
     #[must_use = "sender must be used and notified on block validation success"]
-    pub(super) fn terminate_caching(
+    pub fn terminate_caching(
         &mut self,
         execution_outcome: Option<Arc<BlockExecutionOutput<R>>>,
     ) -> Option<mpsc::Sender<()>> {
@@ -778,7 +840,7 @@ impl<R> Drop for CacheTaskHandle<R> {
 /// - Prepares data for state root proof computation
 /// - Runs concurrently but must not interfere with cache saves
 #[derive(Clone, Debug, Default)]
-struct PayloadExecutionCache {
+pub struct PayloadExecutionCache {
     /// Guarded cloneable cache identified by a block hash.
     inner: Arc<RwLock<Option<SavedCache>>>,
     /// Metrics for cache operations.
@@ -858,7 +920,7 @@ impl PayloadExecutionCache {
     ///
     /// Violating this requirement can result in cache corruption, incorrect state data,
     /// and potential consensus failures.
-    pub(crate) fn update_with_guard<F>(&self, update_fn: F)
+    pub fn update_with_guard<F>(&self, update_fn: F)
     where
         F: FnOnce(&mut Option<SavedCache>),
     {
@@ -885,6 +947,14 @@ pub struct ExecutionEnv<Evm: ConfigureEvm> {
     pub hash: B256,
     /// Hash of the parent block.
     pub parent_hash: B256,
+    /// State root of the parent block.
+    /// Used for sparse trie continuation: if the preserved trie's anchor matches this,
+    /// the trie can be reused directly.
+    pub parent_state_root: B256,
+    /// Number of transactions in the block.
+    /// Used to determine parallel worker count for prewarming.
+    /// A value of 0 indicates the count is unknown.
+    pub transaction_count: usize,
 }
 
 impl<Evm: ConfigureEvm> Default for ExecutionEnv<Evm>
@@ -896,6 +966,8 @@ where
             evm_env: Default::default(),
             hash: Default::default(),
             parent_hash: Default::default(),
+            parent_state_root: Default::default(),
+            transaction_count: 0,
         }
     }
 }

crates/engine/tree/src/tree/payload_processor/multiproof.rs

@@ -100,7 +100,7 @@ impl SparseTrieUpdate {
 
 /// Messages used internally by the multi proof task.
 #[derive(Debug)]
-pub(super) enum MultiProofMessage {
+pub enum MultiProofMessage {
     /// Prefetch proof targets
     PrefetchProofs(VersionedMultiProofTargets),
     /// New state update from transaction execution with its source
@@ -257,7 +257,7 @@ fn extend_multiproof_targets(dest: &mut MultiProofTargets, src: &VersionedMultiP
 
 /// A set of multiproof targets which can be either in the legacy or V2 representations.
 #[derive(Debug)]
-pub(super) enum VersionedMultiProofTargets {
+pub enum VersionedMultiProofTargets {
     /// Legacy targets
     Legacy(MultiProofTargets),
     /// V2 targets
@@ -587,6 +587,8 @@ pub(crate) struct MultiProofTaskMetrics {
     pub first_update_wait_time_histogram: Histogram,
     /// Total time spent waiting for the last proof result.
     pub last_proof_wait_time_histogram: Histogram,
+    /// Time spent preparing the sparse trie for reuse after state root computation.
+    pub into_trie_for_reuse_duration_histogram: Histogram,
 }
 
 /// Standalone task that receives a transaction state stream and updates relevant

crates/engine/tree/src/tree/payload_processor/preserved_sparse_trie.rs

@@ -0,0 +1,117 @@
+//! Preserved sparse trie for reuse across payload validations.
+
+use alloy_primitives::B256;
+use parking_lot::Mutex;
+use reth_trie_sparse::SparseStateTrie;
+use reth_trie_sparse_parallel::ParallelSparseTrie;
+use std::sync::Arc;
+use tracing::debug;
+
+/// Type alias for the sparse trie type used in preservation.
+pub(super) type SparseTrie = SparseStateTrie<ParallelSparseTrie, ParallelSparseTrie>;
+
+/// Shared handle to a preserved sparse trie that can be reused across payload validations.
+///
+/// This is stored in [`PayloadProcessor`](super::PayloadProcessor) and cloned to pass to
+/// [`SparseTrieTask`](super::sparse_trie::SparseTrieTask) for trie reuse.
+#[derive(Debug, Default, Clone)]
+pub(super) struct SharedPreservedSparseTrie(Arc<Mutex<Option<PreservedSparseTrie>>>);
+
+impl SharedPreservedSparseTrie {
+    /// Takes the preserved trie if present, leaving `None` in its place.
+    pub(super) fn take(&self) -> Option<PreservedSparseTrie> {
+        self.0.lock().take()
+    }
+
+    /// Acquires a guard that blocks `take()` until dropped.
+    /// Use this before sending the state root result to ensure the next block
+    /// waits for the trie to be stored.
+    pub(super) fn lock(&self) -> PreservedTrieGuard<'_> {
+        PreservedTrieGuard(self.0.lock())
+    }
+}
+
+/// Guard that holds the lock on the preserved trie.
+/// While held, `take()` will block. Call `store()` to save the trie before dropping.
+pub(super) struct PreservedTrieGuard<'a>(parking_lot::MutexGuard<'a, Option<PreservedSparseTrie>>);
+
+impl PreservedTrieGuard<'_> {
+    /// Stores a preserved trie for later reuse.
+    pub(super) fn store(&mut self, trie: PreservedSparseTrie) {
+        self.0.replace(trie);
+    }
+}
+
+/// A preserved sparse trie that can be reused across payload validations.
+///
+/// The trie exists in one of two states:
+/// - **Anchored**: Has a computed state root and can be reused for payloads whose parent state root
+///   matches the anchor.
+/// - **Cleared**: Trie data has been cleared but allocations are preserved for reuse.
+#[derive(Debug)]
+pub(super) enum PreservedSparseTrie {
+    /// Trie with a computed state root that can be reused for continuation payloads.
+    Anchored {
+        /// The sparse state trie (pruned after root computation).
+        trie: SparseTrie,
+        /// The state root this trie represents (computed from the previous block).
+        /// Used to verify continuity: new payload's `parent_state_root` must match this.
+        state_root: B256,
+    },
+    /// Cleared trie with preserved allocations, ready for fresh use.
+    Cleared {
+        /// The sparse state trie with cleared data but preserved allocations.
+        trie: SparseTrie,
+    },
+}
+
+impl PreservedSparseTrie {
+    /// Creates a new anchored preserved trie.
+    ///
+    /// The `state_root` is the computed state root from the trie, which becomes the
+    /// anchor for determining if subsequent payloads can reuse this trie.
+    pub(super) const fn anchored(trie: SparseTrie, state_root: B256) -> Self {
+        Self::Anchored { trie, state_root }
+    }
+
+    /// Creates a cleared preserved trie (allocations preserved, data cleared).
+    pub(super) const fn cleared(trie: SparseTrie) -> Self {
+        Self::Cleared { trie }
+    }
+
+    /// Consumes self and returns the trie for reuse.
+    ///
+    /// If the preserved trie is anchored and the parent state root matches, the pruned
+    /// trie structure is reused directly. Otherwise, the trie is cleared but allocations
+    /// are preserved to reduce memory overhead.
+    pub(super) fn into_trie_for(self, parent_state_root: B256) -> SparseTrie {
+        match self {
+            Self::Anchored { trie, state_root } if state_root == parent_state_root => {
+                debug!(
+                    target: "engine::tree::payload_processor",
+                    %state_root,
+                    "Reusing anchored sparse trie for continuation payload"
+                );
+                trie
+            }
+            Self::Anchored { mut trie, state_root } => {
+                debug!(
+                    target: "engine::tree::payload_processor",
+                    anchor_root = %state_root,
+                    %parent_state_root,
+                    "Clearing anchored sparse trie - parent state root mismatch"
+                );
+                trie.clear();
+                trie
+            }
+            Self::Cleared { trie } => {
+                debug!(
+                    target: "engine::tree::payload_processor",
+                    %parent_state_root,
+                    "Using cleared sparse trie with preserved allocations"
+                );
+                trie
+            }
+        }
+    }
+}

crates/engine/tree/src/tree/payload_processor/prewarm.rs

@@ -49,7 +49,8 @@ use std::{
 use tracing::{debug, debug_span, instrument, trace, warn, Span};
 
 /// Determines the prewarming mode: transaction-based or BAL-based.
-pub(super) enum PrewarmMode<Tx> {
+#[derive(Debug)]
+pub enum PrewarmMode<Tx> {
     /// Prewarm by executing transactions from a stream.
     Transactions(Receiver<Tx>),
     /// Prewarm by prefetching slots from a Block Access List.
@@ -69,7 +70,8 @@ struct IndexedTransaction<Tx> {
 /// individually in parallel.
 ///
 /// Note: This task runs until cancelled externally.
-pub(super) struct PrewarmCacheTask<N, P, Evm>
+#[derive(Debug)]
+pub struct PrewarmCacheTask<N, P, Evm>
 where
     N: NodePrimitives,
     Evm: ConfigureEvm<Primitives = N>,
@@ -82,8 +84,6 @@ where
     ctx: PrewarmContext<N, P, Evm>,
     /// How many transactions should be executed in parallel
     max_concurrency: usize,
-    /// The number of transactions to be processed
-    transaction_count_hint: usize,
     /// Sender to emit evm state outcome messages, if any.
     to_multi_proof: Option<CrossbeamSender<MultiProofMessage>>,
     /// Receiver for events produced by tx execution
@@ -99,12 +99,11 @@ where
     Evm: ConfigureEvm<Primitives = N> + 'static,
 {
     /// Initializes the task with the given transactions pending execution
-    pub(super) fn new(
+    pub fn new(
         executor: WorkloadExecutor,
         execution_cache: PayloadExecutionCache,
         ctx: PrewarmContext<N, P, Evm>,
         to_multi_proof: Option<CrossbeamSender<MultiProofMessage>>,
-        transaction_count_hint: usize,
         max_concurrency: usize,
     ) -> (Self, Sender<PrewarmTaskEvent<N::Receipt>>) {
         let (actions_tx, actions_rx) = channel();
@@ -112,7 +111,7 @@ where
         trace!(
             target: "engine::tree::payload_processor::prewarm",
             max_concurrency,
-            transaction_count_hint,
+            transaction_count = ctx.env.transaction_count,
             "Initialized prewarm task"
         );
 
@@ -122,7 +121,6 @@ where
                 execution_cache,
                 ctx,
                 max_concurrency,
-                transaction_count_hint,
                 to_multi_proof,
                 actions_rx,
                 parent_span: Span::current(),
@@ -146,7 +144,6 @@ where
         let executor = self.executor.clone();
         let ctx = self.ctx.clone();
         let max_concurrency = self.max_concurrency;
-        let transaction_count_hint = self.transaction_count_hint;
         let span = Span::current();
 
         self.executor.spawn_blocking(move || {
@@ -154,13 +151,14 @@ where
 
             let (done_tx, done_rx) = mpsc::channel();
 
-            // When transaction_count_hint is 0, it means the count is unknown. In this case, spawn
+            // When transaction_count is 0, it means the count is unknown. In this case, spawn
             // max workers to handle potentially many transactions in parallel rather
             // than bottlenecking on a single worker.
-            let workers_needed = if transaction_count_hint == 0 {
+            let transaction_count = ctx.env.transaction_count;
+            let workers_needed = if transaction_count == 0 {
                 max_concurrency
             } else {
-                transaction_count_hint.min(max_concurrency)
+                transaction_count.min(max_concurrency)
             };
 
             // Spawn workers
@@ -370,11 +368,8 @@ where
         name = "prewarm and caching",
         skip_all
     )]
-    pub(super) fn run<Tx>(
-        self,
-        mode: PrewarmMode<Tx>,
-        actions_tx: Sender<PrewarmTaskEvent<N::Receipt>>,
-    ) where
+    pub fn run<Tx>(self, mode: PrewarmMode<Tx>, actions_tx: Sender<PrewarmTaskEvent<N::Receipt>>)
+    where
         Tx: ExecutableTxFor<Evm> + Clone + Send + 'static,
     {
         // Spawn execution tasks based on mode
@@ -436,23 +431,29 @@ where
 
 /// Context required by tx execution tasks.
 #[derive(Debug, Clone)]
-pub(super) struct PrewarmContext<N, P, Evm>
+pub struct PrewarmContext<N, P, Evm>
 where
     N: NodePrimitives,
     Evm: ConfigureEvm<Primitives = N>,
 {
-    pub(super) env: ExecutionEnv<Evm>,
-    pub(super) evm_config: Evm,
-    pub(super) saved_cache: Option<SavedCache>,
+    /// The execution environment.
+    pub env: ExecutionEnv<Evm>,
+    /// The EVM configuration.
+    pub evm_config: Evm,
+    /// The saved cache.
+    pub saved_cache: Option<SavedCache>,
     /// Provider to obtain the state
-    pub(super) provider: StateProviderBuilder<N, P>,
-    pub(super) metrics: PrewarmMetrics,
+    pub provider: StateProviderBuilder<N, P>,
+    /// The metrics for the prewarm task.
+    pub metrics: PrewarmMetrics,
     /// An atomic bool that tells prewarm tasks to not start any more execution.
-    pub(super) terminate_execution: Arc<AtomicBool>,
-    pub(super) precompile_cache_disabled: bool,
-    pub(super) precompile_cache_map: PrecompileCacheMap<SpecFor<Evm>>,
+    pub terminate_execution: Arc<AtomicBool>,
+    /// Whether the precompile cache is disabled.
+    pub precompile_cache_disabled: bool,
+    /// The precompile cache map.
+    pub precompile_cache_map: PrecompileCacheMap<SpecFor<Evm>>,
     /// Whether V2 proof calculation is enabled.
-    pub(super) v2_proofs_enabled: bool,
+    pub v2_proofs_enabled: bool,
 }
 
 impl<N, P, Evm> PrewarmContext<N, P, Evm>
@@ -563,6 +564,7 @@ where
                 index,
                 tx_hash = %tx.tx().tx_hash(),
                 is_success = tracing::field::Empty,
+                gas_used = tracing::field::Empty,
             )
             .entered();
 
@@ -851,7 +853,8 @@ fn multiproof_targets_v2_from_state(state: EvmState) -> (VersionedMultiProofTarg
 ///
 /// Generic over `R` (receipt type) to allow sharing `Arc<ExecutionOutcome<R>>` with the main
 /// execution path without cloning the expensive `BundleState`.
-pub(super) enum PrewarmTaskEvent<R> {
+#[derive(Debug)]
+pub enum PrewarmTaskEvent<R> {
     /// Forcefully terminate all remaining transaction execution.
     TerminateTransactionExecution,
     /// Forcefully terminate the task on demand and update the shared cache with the given output
@@ -881,7 +884,7 @@ pub(super) enum PrewarmTaskEvent<R> {
 /// Metrics for transactions prewarming.
 #[derive(Metrics, Clone)]
 #[metrics(scope = "sync.prewarm")]
-pub(crate) struct PrewarmMetrics {
+pub struct PrewarmMetrics {
     /// The number of transactions to prewarm
     pub(crate) transactions: Gauge,
     /// A histogram of the number of transactions to prewarm

crates/engine/tree/src/tree/payload_processor/receipt_root_task.rs

@@ -77,8 +77,22 @@ impl<R: Receipt> ReceiptRootTaskHandle<R> {
             receipt_with_bloom.encode_2718(&mut encode_buf);
 
             aggregated_bloom |= *receipt_with_bloom.bloom_ref();
-            builder.push_unchecked(indexed_receipt.index, &encode_buf);
-            received_count += 1;
+            match builder.push(indexed_receipt.index, &encode_buf) {
+                Ok(()) => {
+                    received_count += 1;
+                }
+                Err(err) => {
+                    // If a duplicate or out-of-bounds index is streamed, skip it and
+                    // fall back to computing the receipt root from the full receipts
+                    // vector later.
+                    tracing::error!(
+                        target: "engine::tree::payload_processor",
+                        index = indexed_receipt.index,
+                        ?err,
+                        "Receipt root task received invalid receipt index, skipping"
+                    );
+                }
+            }
         }
 
         let Ok(root) = builder.finalize() else {

crates/engine/tree/src/tree/payload_processor/sparse_trie.rs

@@ -1,15 +1,34 @@
 //! Sparse Trie task related functionality.
 
-use crate::tree::payload_processor::multiproof::{MultiProofTaskMetrics, SparseTrieUpdate};
+use crate::tree::{
+    multiproof::{evm_state_to_hashed_post_state, MultiProofMessage, VersionedMultiProofTargets},
+    payload_processor::multiproof::{MultiProofTaskMetrics, SparseTrieUpdate},
+};
 use alloy_primitives::B256;
-use rayon::iter::{ParallelBridge, ParallelIterator};
-use reth_trie::{updates::TrieUpdates, Nibbles};
-use reth_trie_parallel::{proof_task::ProofResult, root::ParallelStateRootError};
+use alloy_rlp::{Decodable, Encodable};
+use crossbeam_channel::{Receiver as CrossbeamReceiver, Sender as CrossbeamSender};
+use rayon::iter::ParallelIterator;
+use reth_errors::ProviderError;
+use reth_primitives_traits::{Account, ParallelBridgeBuffered};
+use reth_revm::state::EvmState;
+use reth_trie::{
+    proof_v2::Target, updates::TrieUpdates, HashedPostState, Nibbles, TrieAccount, EMPTY_ROOT_HASH,
+    TRIE_ACCOUNT_RLP_MAX_SIZE,
+};
+use reth_trie_parallel::{
+    proof_task::{
+        AccountMultiproofInput, ProofResult, ProofResultContext, ProofResultMessage,
+        ProofWorkerHandle,
+    },
+    root::ParallelStateRootError,
+    targets_v2::MultiProofTargetsV2,
+};
 use reth_trie_sparse::{
     errors::{SparseStateTrieResult, SparseTrieErrorKind},
     provider::{TrieNodeProvider, TrieNodeProviderFactory},
-    ClearedSparseStateTrie, SerialSparseTrie, SparseStateTrie, SparseTrie,
+    LeafUpdate, SerialSparseTrie, SparseStateTrie, SparseTrie, SparseTrieExt,
 };
+use revm_primitives::{hash_map::Entry, B256Map};
 use smallvec::SmallVec;
 use std::{
     sync::mpsc,
@@ -17,6 +36,64 @@ use std::{
 };
 use tracing::{debug, debug_span, instrument, trace};
 
+#[expect(clippy::large_enum_variant)]
+pub(super) enum SpawnedSparseTrieTask<BPF, A, S>
+where
+    BPF: TrieNodeProviderFactory + Send + Sync,
+    BPF::AccountNodeProvider: TrieNodeProvider + Send + Sync,
+    BPF::StorageNodeProvider: TrieNodeProvider + Send + Sync,
+    A: SparseTrie + SparseTrieExt + Send + Sync + Default,
+    S: SparseTrie + SparseTrieExt + Send + Sync + Default + Clone,
+{
+    Cleared(SparseTrieTask<BPF, A, S>),
+    Cached(SparseTrieCacheTask<A, S>),
+}
+
+impl<BPF, A, S> SpawnedSparseTrieTask<BPF, A, S>
+where
+    BPF: TrieNodeProviderFactory + Send + Sync + Clone,
+    BPF::AccountNodeProvider: TrieNodeProvider + Send + Sync,
+    BPF::StorageNodeProvider: TrieNodeProvider + Send + Sync,
+    A: SparseTrie + SparseTrieExt + Send + Sync + Default,
+    S: SparseTrie + SparseTrieExt + Send + Sync + Default + Clone,
+{
+    pub(super) fn run(&mut self) -> Result<StateRootComputeOutcome, ParallelStateRootError> {
+        match self {
+            Self::Cleared(task) => task.run(),
+            Self::Cached(task) => task.run(),
+        }
+    }
+
+    pub(super) fn into_trie_for_reuse(
+        self,
+        prune_depth: usize,
+        max_storage_tries: usize,
+        max_nodes_capacity: usize,
+        max_values_capacity: usize,
+    ) -> SparseStateTrie<A, S> {
+        match self {
+            Self::Cleared(task) => task.into_cleared_trie(max_nodes_capacity, max_values_capacity),
+            Self::Cached(task) => task.into_trie_for_reuse(
+                prune_depth,
+                max_storage_tries,
+                max_nodes_capacity,
+                max_values_capacity,
+            ),
+        }
+    }
+
+    pub(super) fn into_cleared_trie(
+        self,
+        max_nodes_capacity: usize,
+        max_values_capacity: usize,
+    ) -> SparseStateTrie<A, S> {
+        match self {
+            Self::Cleared(task) => task.into_cleared_trie(max_nodes_capacity, max_values_capacity),
+            Self::Cached(task) => task.into_cleared_trie(max_nodes_capacity, max_values_capacity),
+        }
+    }
+}
+
 /// A task responsible for populating the sparse trie.
 pub(super) struct SparseTrieTask<BPF, A = SerialSparseTrie, S = SerialSparseTrie>
 where
@@ -38,46 +115,29 @@ where
     BPF: TrieNodeProviderFactory + Send + Sync + Clone,
     BPF::AccountNodeProvider: TrieNodeProvider + Send + Sync,
     BPF::StorageNodeProvider: TrieNodeProvider + Send + Sync,
-    A: SparseTrie + Send + Sync + Default,
-    S: SparseTrie + Send + Sync + Default + Clone,
+    A: SparseTrie + SparseTrieExt + Send + Sync + Default,
+    S: SparseTrie + SparseTrieExt + Send + Sync + Default + Clone,
 {
-    /// Creates a new sparse trie, pre-populating with a [`ClearedSparseStateTrie`].
-    pub(super) fn new_with_cleared_trie(
+    /// Creates a new sparse trie task with the given trie.
+    pub(super) const fn new(
         updates: mpsc::Receiver<SparseTrieUpdate>,
         blinded_provider_factory: BPF,
         metrics: MultiProofTaskMetrics,
-        sparse_state_trie: ClearedSparseStateTrie<A, S>,
+        trie: SparseStateTrie<A, S>,
     ) -> Self {
-        Self { updates, metrics, trie: sparse_state_trie.into_inner(), blinded_provider_factory }
+        Self { updates, metrics, trie, blinded_provider_factory }
     }
 
-    /// Runs the sparse trie task to completion.
+    /// Runs the sparse trie task to completion, computing the state root.
     ///
-    /// This waits for new incoming [`SparseTrieUpdate`].
-    ///
-    /// This concludes once the last trie update has been received.
-    ///
-    /// # Returns
-    ///
-    /// - State root computation outcome.
-    /// - `SparseStateTrie` that needs to be cleared and reused to avoid reallocations.
+    /// Receives [`SparseTrieUpdate`]s until the channel is closed, applying each update
+    /// to the trie. Once all updates are processed, computes and returns the final state root.
     #[instrument(
         level = "debug",
         target = "engine::tree::payload_processor::sparse_trie",
         skip_all
     )]
-    pub(super) fn run(
-        mut self,
-    ) -> (Result<StateRootComputeOutcome, ParallelStateRootError>, SparseStateTrie<A, S>) {
-        // run the main loop to completion
-        let result = self.run_inner();
-        (result, self.trie)
-    }
-
-    /// Inner function to run the sparse trie task to completion.
-    ///
-    /// See [`Self::run`] for more information.
-    fn run_inner(&mut self) -> Result<StateRootComputeOutcome, ParallelStateRootError> {
+    pub(super) fn run(&mut self) -> Result<StateRootComputeOutcome, ParallelStateRootError> {
         let now = Instant::now();
 
         let mut num_iterations = 0;
@@ -127,6 +187,433 @@ where
 
         Ok(StateRootComputeOutcome { state_root, trie_updates })
     }
+
+    /// Clears and shrinks the trie, discarding all state.
+    ///
+    /// Use this when the payload was invalid or cancelled - we don't want to preserve
+    /// potentially invalid trie state, but we keep the allocations for reuse.
+    pub(super) fn into_cleared_trie(
+        mut self,
+        max_nodes_capacity: usize,
+        max_values_capacity: usize,
+    ) -> SparseStateTrie<A, S> {
+        self.trie.clear();
+        self.trie.shrink_to(max_nodes_capacity, max_values_capacity);
+        self.trie
+    }
+}
+
+/// Sparse trie task implementation that uses in-memory sparse trie data to schedule proof fetching.
+pub(super) struct SparseTrieCacheTask<A = SerialSparseTrie, S = SerialSparseTrie> {
+    /// Sender for proof results.
+    proof_result_tx: CrossbeamSender<ProofResultMessage>,
+    /// Receiver for proof results directly from workers.
+    proof_result_rx: CrossbeamReceiver<ProofResultMessage>,
+    /// Receives updates from execution and prewarming.
+    updates: CrossbeamReceiver<MultiProofMessage>,
+    /// `SparseStateTrie` used for computing the state root.
+    trie: SparseStateTrie<A, S>,
+    /// Handle to the proof worker pools (storage and account).
+    proof_worker_handle: ProofWorkerHandle,
+    /// Account trie updates.
+    account_updates: B256Map<LeafUpdate>,
+    /// Storage trie updates. hashed address -> slot -> update.
+    storage_updates: B256Map<B256Map<LeafUpdate>>,
+    /// Account updates that are blocked by storage root calculation or account reveal.
+    ///
+    /// Those are being moved into `account_updates` once storage roots
+    /// are revealed and/or calculated.
+    ///
+    /// Invariant: for each entry in `pending_account_updates` account must either be already
+    /// revealed in the trie or have an entry in `account_updates`.
+    ///
+    /// Values can be either of:
+    ///   - None: account had a storage update and is awaiting storage root calculation and/or
+    ///     account node reveal to complete.
+    ///   - Some(_): account was changed/destroyed and is awaiting storage root calculation/reveal
+    ///     to complete.
+    pending_account_updates: B256Map<Option<Option<Account>>>,
+    /// Cache of account proof targets that were already fetched/requested from the proof workers.
+    /// account -> lowest `min_len` requested.
+    fetched_account_targets: B256Map<u8>,
+    /// Cache of storage proof targets that have already been fetched/requested from the proof
+    /// workers. account -> slot -> lowest `min_len` requested.
+    fetched_storage_targets: B256Map<B256Map<u8>>,
+    /// Reusable buffer for RLP encoding of accounts.
+    account_rlp_buf: Vec<u8>,
+    /// Metrics for the sparse trie.
+    metrics: MultiProofTaskMetrics,
+}
+
+impl<A, S> SparseTrieCacheTask<A, S>
+where
+    A: SparseTrieExt + Default,
+    S: SparseTrieExt + Default + Clone,
+{
+    /// Creates a new sparse trie, pre-populating with an existing [`SparseStateTrie`].
+    pub(super) fn new_with_trie(
+        updates: CrossbeamReceiver<MultiProofMessage>,
+        proof_worker_handle: ProofWorkerHandle,
+        metrics: MultiProofTaskMetrics,
+        sparse_state_trie: SparseStateTrie<A, S>,
+    ) -> Self {
+        let (proof_result_tx, proof_result_rx) = crossbeam_channel::unbounded();
+        Self {
+            proof_result_tx,
+            proof_result_rx,
+            updates,
+            proof_worker_handle,
+            trie: sparse_state_trie,
+            account_updates: Default::default(),
+            storage_updates: Default::default(),
+            pending_account_updates: Default::default(),
+            fetched_account_targets: Default::default(),
+            fetched_storage_targets: Default::default(),
+            account_rlp_buf: Vec::with_capacity(TRIE_ACCOUNT_RLP_MAX_SIZE),
+            metrics,
+        }
+    }
+
+    /// Prunes and shrinks the trie for reuse in the next payload built on top of this one.
+    ///
+    /// Should be called after the state root result has been sent.
+    pub(super) fn into_trie_for_reuse(
+        mut self,
+        prune_depth: usize,
+        max_storage_tries: usize,
+        max_nodes_capacity: usize,
+        max_values_capacity: usize,
+    ) -> SparseStateTrie<A, S> {
+        self.trie.prune(prune_depth, max_storage_tries);
+        self.trie.shrink_to(max_nodes_capacity, max_values_capacity);
+        self.trie
+    }
+
+    /// Clears and shrinks the trie, discarding all state.
+    ///
+    /// Use this when the payload was invalid or cancelled - we don't want to preserve
+    /// potentially invalid trie state, but we keep the allocations for reuse.
+    pub(super) fn into_cleared_trie(
+        mut self,
+        max_nodes_capacity: usize,
+        max_values_capacity: usize,
+    ) -> SparseStateTrie<A, S> {
+        self.trie.clear();
+        self.trie.shrink_to(max_nodes_capacity, max_values_capacity);
+        self.trie
+    }
+
+    /// Runs the sparse trie task to completion.
+    ///
+    /// This waits for new incoming [`MultiProofMessage`]s, applies updates to the trie and
+    /// schedules proof fetching when needed.
+    ///
+    /// This concludes once the last state update has been received and processed.
+    #[instrument(
+        level = "debug",
+        target = "engine::tree::payload_processor::sparse_trie",
+        skip_all
+    )]
+    pub(super) fn run(&mut self) -> Result<StateRootComputeOutcome, ParallelStateRootError> {
+        let now = Instant::now();
+
+        let mut finished_state_updates = false;
+        loop {
+            crossbeam_channel::select_biased! {
+                recv(self.proof_result_rx) -> message => {
+                    let Ok(result) = message else {
+                        unreachable!("we own the sender half")
+                    };
+                    self.on_proof_result(result)?;
+                },
+                recv(self.updates) -> message => {
+                    let update = match message {
+                        Ok(m) => m,
+                        Err(_) => {
+                            break
+                        }
+                    };
+
+                    match update {
+                        MultiProofMessage::PrefetchProofs(targets) => {
+                            self.on_prewarm_targets(targets);
+                        }
+                        MultiProofMessage::StateUpdate(_, state) => {
+                            self.on_state_update(state);
+                        }
+                        MultiProofMessage::EmptyProof { sequence_number: _, state } => {
+                            self.on_hashed_state_update(state);
+                        }
+                        MultiProofMessage::BlockAccessList(_) => todo!(),
+                        MultiProofMessage::FinishedStateUpdates => {
+                            finished_state_updates = true;
+                        }
+                    }
+                }
+            }
+
+            self.process_updates()?;
+
+            if finished_state_updates &&
+                self.account_updates.is_empty() &&
+                self.storage_updates.iter().all(|(_, updates)| updates.is_empty())
+            {
+                break;
+            }
+        }
+
+        debug!(target: "engine::root", "All proofs processed, ending calculation");
+
+        let start = Instant::now();
+        let (state_root, trie_updates) =
+            self.trie.root_with_updates(&self.proof_worker_handle).map_err(|e| {
+                ParallelStateRootError::Other(format!("could not calculate state root: {e:?}"))
+            })?;
+
+        let end = Instant::now();
+        self.metrics.sparse_trie_final_update_duration_histogram.record(end.duration_since(start));
+        self.metrics.sparse_trie_total_duration_histogram.record(end.duration_since(now));
+
+        Ok(StateRootComputeOutcome { state_root, trie_updates })
+    }
+
+    fn on_prewarm_targets(&mut self, targets: VersionedMultiProofTargets) {
+        let VersionedMultiProofTargets::V2(targets) = targets else {
+            unreachable!("sparse trie as cache must only be used with V2 multiproof targets");
+        };
+
+        for target in targets.account_targets {
+            // Only touch accounts that are not yet present in the updates set.
+            self.account_updates.entry(target.key()).or_insert(LeafUpdate::Touched);
+        }
+
+        for (address, slots) in targets.storage_targets {
+            for slot in slots {
+                // Only touch storages that are not yet present in the updates set.
+                self.storage_updates
+                    .entry(address)
+                    .or_default()
+                    .entry(slot.key())
+                    .or_insert(LeafUpdate::Touched);
+            }
+
+            // Touch corresponding account leaf to make sure its revealed in accounts trie for
+            // storage root update.
+            self.account_updates.entry(address).or_insert(LeafUpdate::Touched);
+        }
+    }
+
+    /// Processes a state update and encodes all state changes as trie updates.
+    #[instrument(
+        level = "debug",
+        target = "engine::tree::payload_processor::sparse_trie",
+        skip_all,
+        fields(accounts = update.len())
+    )]
+    fn on_state_update(&mut self, update: EvmState) {
+        let hashed_state_update = evm_state_to_hashed_post_state(update);
+        self.on_hashed_state_update(hashed_state_update)
+    }
+
+    /// Processes a hashed state update and encodes all state changes as trie updates.
+    fn on_hashed_state_update(&mut self, hashed_state_update: HashedPostState) {
+        for (address, storage) in hashed_state_update.storages {
+            for (slot, value) in storage.storage {
+                let encoded = if value.is_zero() {
+                    Vec::new()
+                } else {
+                    alloy_rlp::encode_fixed_size(&value).to_vec()
+                };
+                self.storage_updates
+                    .entry(address)
+                    .or_default()
+                    .insert(slot, LeafUpdate::Changed(encoded));
+            }
+
+            // Make sure account is tracked in `account_updates` so that it is revealed in accounts
+            // trie for storage root update.
+            self.account_updates.entry(address).or_insert(LeafUpdate::Touched);
+
+            // Make sure account is tracked in `pending_account_updates` so that once storage root
+            // is computed, it will be updated in the accounts trie.
+            self.pending_account_updates.entry(address).or_insert(None);
+        }
+
+        for (address, account) in hashed_state_update.accounts {
+            // Track account as touched.
+            //
+            // This might overwrite an existing update, which is fine, because storage root from it
+            // is already tracked in the trie and can be easily fetched again.
+            self.account_updates.insert(address, LeafUpdate::Touched);
+
+            // Track account in `pending_account_updates` so that once storage root is computed,
+            // it will be updated in the accounts trie.
+            self.pending_account_updates.insert(address, Some(account));
+        }
+    }
+
+    fn on_proof_result(
+        &mut self,
+        result: ProofResultMessage,
+    ) -> Result<(), ParallelStateRootError> {
+        let ProofResult::V2(result) = result.result? else {
+            unreachable!("sparse trie as cache must only be used with multiproof v2");
+        };
+
+        self.trie.reveal_decoded_multiproof_v2(result).map_err(|e| {
+            ParallelStateRootError::Other(format!("could not reveal multiproof: {e:?}"))
+        })
+    }
+
+    /// Applies updates to the sparse trie and dispatches requested multiproof targets.
+    fn process_updates(&mut self) -> Result<(), ProviderError> {
+        let mut targets = MultiProofTargetsV2::default();
+
+        for (addr, updates) in &mut self.storage_updates {
+            let trie = self.trie.get_or_create_storage_trie_mut(*addr);
+            let fetched_storage = self.fetched_storage_targets.entry(*addr).or_default();
+
+            trie.update_leaves(updates, |path, min_len| match fetched_storage.entry(path) {
+                Entry::Occupied(mut entry) => {
+                    if min_len < *entry.get() {
+                        entry.insert(min_len);
+                        targets
+                            .storage_targets
+                            .entry(*addr)
+                            .or_default()
+                            .push(Target::new(path).with_min_len(min_len));
+                    }
+                }
+                Entry::Vacant(entry) => {
+                    entry.insert(min_len);
+                    targets
+                        .storage_targets
+                        .entry(*addr)
+                        .or_default()
+                        .push(Target::new(path).with_min_len(min_len));
+                }
+            })
+            .map_err(ProviderError::other)?;
+
+            // If all storage updates were processed, we can now compute the new storage root.
+            if updates.is_empty() {
+                let storage_root =
+                    trie.root().expect("updates are drained, trie should be revealed by now");
+
+                // If there is a pending account update for this address with known info, we can
+                // encode it into proper update right away.
+                if let Entry::Occupied(entry) = self.pending_account_updates.entry(*addr) &&
+                    entry.get().is_some()
+                {
+                    let account = entry.remove().expect("just checked, should be Some");
+                    let encoded = if account.is_none_or(|account| account.is_empty()) &&
+                        storage_root == EMPTY_ROOT_HASH
+                    {
+                        Vec::new()
+                    } else {
+                        self.account_rlp_buf.clear();
+                        account
+                            .unwrap_or_default()
+                            .into_trie_account(storage_root)
+                            .encode(&mut self.account_rlp_buf);
+                        self.account_rlp_buf.clone()
+                    };
+                    self.account_updates.insert(*addr, LeafUpdate::Changed(encoded));
+                }
+            }
+        }
+
+        loop {
+            // Now handle pending account updates that can be upgraded to a proper update.
+            let account_rlp_buf = &mut self.account_rlp_buf;
+            self.pending_account_updates.retain(|addr, account| {
+                // If account has pending storage updates, it is still pending.
+                if self.storage_updates.get(addr).is_some_and(|updates| !updates.is_empty()) {
+                    return true;
+                }
+
+                // Get the current account state either from the trie or from latest account update.
+                let trie_account = if let Some(LeafUpdate::Changed(encoded)) = self.account_updates.get(addr) {
+                    Some(encoded).filter(|encoded| !encoded.is_empty())
+                } else if !self.account_updates.contains_key(addr) {
+                    self.trie.get_account_value(addr)
+                } else {
+                    // Needs to be revealed first
+                    return true;
+                };
+
+                let trie_account = trie_account.map(|value| TrieAccount::decode(&mut &value[..]).expect("invalid account RLP"));
+
+                let (account, storage_root) = if let Some(account) = account.take() {
+                    // If account is Some(_) here it means it didn't have any storage updates
+                    // and we can fetch the storage root directly from the account trie.
+                    //
+                    // If it did have storage updates, we would've had processed it above when iterating over storage tries.
+                    let storage_root = trie_account.map(|account| account.storage_root).unwrap_or(EMPTY_ROOT_HASH);
+
+                    (account, storage_root)
+                } else {
+                    (trie_account.map(Into::into), self.trie.storage_root(addr).expect("account had storage updates that were applied to its trie, storage root must be revealed by now"))
+                };
+
+                let encoded = if account.is_none_or(|account| account.is_empty()) && storage_root == EMPTY_ROOT_HASH {
+                    Vec::new()
+                } else {
+                    account_rlp_buf.clear();
+                    account.unwrap_or_default().into_trie_account(storage_root).encode(account_rlp_buf);
+                    account_rlp_buf.clone()
+                };
+                self.account_updates.insert(*addr, LeafUpdate::Changed(encoded));
+
+                false
+            });
+
+            let updates_len_before = self.account_updates.len();
+
+            // Process account trie updates and fill the account targets.
+            self.trie
+                .trie_mut()
+                .update_leaves(&mut self.account_updates, |target, min_len| {
+                    match self.fetched_account_targets.entry(target) {
+                        Entry::Occupied(mut entry) => {
+                            if min_len < *entry.get() {
+                                entry.insert(min_len);
+                                targets
+                                    .account_targets
+                                    .push(Target::new(target).with_min_len(min_len));
+                            }
+                        }
+                        Entry::Vacant(entry) => {
+                            entry.insert(min_len);
+                            targets.account_targets.push(Target::new(target).with_min_len(min_len));
+                        }
+                    }
+                })
+                .map_err(ProviderError::other)?;
+
+            if updates_len_before == self.account_updates.len() {
+                // Only exit when no new updates are processed.
+                //
+                // We need to keep iterating if any updates are being drained because that might
+                // indicate that more pending account updates can be promoted.
+                break;
+            }
+        }
+
+        if !targets.is_empty() {
+            self.proof_worker_handle.dispatch_account_multiproof(AccountMultiproofInput::V2 {
+                targets,
+                proof_result_sender: ProofResultContext::new(
+                    self.proof_result_tx.clone(),
+                    0,
+                    HashedPostState::default(),
+                    Instant::now(),
+                ),
+            })?;
+        }
+
+        Ok(())
+    }
 }
 
 /// Outcome of the state root computation, including the state root itself with
@@ -178,7 +665,7 @@ where
         .storages
         .into_iter()
         .map(|(address, storage)| (address, storage, trie.take_storage_trie(&address)))
-        .par_bridge()
+        .par_bridge_buffered()
         .map(|(address, storage, storage_trie)| {
             let _enter =
                 debug_span!(target: "engine::tree::payload_processor::sparse_trie", parent: &span, "storage trie", ?address)

crates/engine/tree/src/tree/payload_validator.rs

@@ -402,7 +402,13 @@ where
             .in_scope(|| self.evm_env_for(&input))
             .map_err(NewPayloadError::other)?;
 
-        let env = ExecutionEnv { evm_env, hash: input.hash(), parent_hash: input.parent_hash() };
+        let env = ExecutionEnv {
+            evm_env,
+            hash: input.hash(),
+            parent_hash: input.parent_hash(),
+            parent_state_root: parent_block.state_root(),
+            transaction_count: input.transaction_count(),
+        };
 
         // Plan the strategy used for state root computation.
         let strategy = self.plan_state_root_computation();
@@ -514,6 +520,14 @@ where
                         info!(target: "engine::tree::payload_validator", ?state_root, ?elapsed, "State root task finished");
                         // we double check the state root here for good measure
                         if state_root == block.header().state_root() {
+                            // Compare trie updates with serial computation if configured
+                            if self.config.always_compare_trie_updates() {
+                                self.compare_trie_updates_with_serial(
+                                    overlay_factory.clone(),
+                                    &hashed_state,
+                                    trie_updates.clone(),
+                                );
+                            }
                             maybe_state_root = Some((state_root, trie_updates, elapsed))
                         } else {
                             warn!(
@@ -792,6 +806,11 @@ where
         // Execute transactions
         let exec_span = debug_span!(target: "engine::tree", "execution").entered();
         let mut transactions = transactions.into_iter();
+        // Some executors may execute transactions that do not append receipts during the
+        // main loop (e.g., system transactions whose receipts are added during finalization).
+        // In that case, invoking the callback on every transaction would resend the previous
+        // receipt with the same index and can panic the ordered root builder.
+        let mut last_sent_len = 0usize;
         loop {
             // Measure time spent waiting for next transaction from iterator
             // (e.g., parallel signature recovery)
@@ -818,10 +837,14 @@ where
             let gas_used = executor.execute_transaction(tx)?;
             self.metrics.record_transaction_execution(tx_start.elapsed());
 
-            // Send the latest receipt to the background task for incremental root computation
-            if let Some(receipt) = executor.receipts().last() {
-                let tx_index = executor.receipts().len() - 1;
-                let _ = receipt_tx.send(IndexedReceipt::new(tx_index, receipt.clone()));
+            let current_len = executor.receipts().len();
+            if current_len > last_sent_len {
+                last_sent_len = current_len;
+                // Send the latest receipt to the background task for incremental root computation.
+                if let Some(receipt) = executor.receipts().last() {
+                    let tx_index = current_len - 1;
+                    let _ = receipt_tx.send(IndexedReceipt::new(tx_index, receipt.clone()));
+                }
             }
 
             enter.record("gas_used", gas_used);
@@ -880,6 +903,62 @@ where
             .root_with_updates()?)
     }
 
+    /// Compares trie updates from the state root task with serial state root computation.
+    ///
+    /// This is used for debugging and validating the correctness of the parallel state root
+    /// task implementation. When enabled via `--engine.state-root-task-compare-updates`, this
+    /// method runs a separate serial state root computation and compares the resulting trie
+    /// updates.
+    fn compare_trie_updates_with_serial(
+        &self,
+        overlay_factory: OverlayStateProviderFactory<P>,
+        hashed_state: &HashedPostState,
+        task_trie_updates: TrieUpdates,
+    ) {
+        debug!(target: "engine::tree::payload_validator", "Comparing trie updates with serial computation");
+
+        match self.compute_state_root_serial(overlay_factory.clone(), hashed_state) {
+            Ok((serial_root, serial_trie_updates)) => {
+                debug!(
+                    target: "engine::tree::payload_validator",
+                    ?serial_root,
+                    "Serial state root computation finished for comparison"
+                );
+
+                // Get a database provider to use as trie cursor factory
+                match overlay_factory.database_provider_ro() {
+                    Ok(provider) => {
+                        if let Err(err) = super::trie_updates::compare_trie_updates(
+                            &provider,
+                            task_trie_updates,
+                            serial_trie_updates,
+                        ) {
+                            warn!(
+                                target: "engine::tree::payload_validator",
+                                %err,
+                                "Error comparing trie updates"
+                            );
+                        }
+                    }
+                    Err(err) => {
+                        warn!(
+                            target: "engine::tree::payload_validator",
+                            %err,
+                            "Failed to get database provider for trie update comparison"
+                        );
+                    }
+                }
+            }
+            Err(err) => {
+                warn!(
+                    target: "engine::tree::payload_validator",
+                    %err,
+                    "Failed to compute serial state root for comparison"
+                );
+            }
+        }
+    }
+
     /// Validates the block after execution.
     ///
     /// This performs:
@@ -1103,10 +1182,13 @@ where
     /// while the trie input computation is deferred until the overlay is actually needed.
     ///
     /// If parent is on disk (no in-memory blocks), returns `None` for the lazy overlay.
+    ///
+    /// Uses a cached overlay if available for the canonical head (the common case).
     fn get_parent_lazy_overlay(
         parent_hash: B256,
         state: &EngineApiTreeState<N>,
     ) -> (Option<LazyOverlay>, B256) {
+        // Get blocks leading to the parent to determine the anchor
         let (anchor_hash, blocks) =
             state.tree_state.blocks_by_hash(parent_hash).unwrap_or_else(|| (parent_hash, vec![]));
 
@@ -1115,6 +1197,17 @@ where
             return (None, anchor_hash);
         }
 
+        // Try to use the cached overlay if it matches both parent hash and anchor
+        if let Some(cached) = state.tree_state.get_cached_overlay(parent_hash, anchor_hash) {
+            debug!(
+                target: "engine::tree::payload_validator",
+                %parent_hash,
+                %anchor_hash,
+                "Using cached canonical overlay"
+            );
+            return (Some(cached.overlay.clone()), cached.anchor_hash);
+        }
+
         debug!(
             target: "engine::tree::payload_validator",
             %anchor_hash,

crates/engine/tree/src/tree/precompile_cache.rs

@@ -21,7 +21,8 @@ impl<S> PrecompileCacheMap<S>
 where
     S: Eq + Hash + std::fmt::Debug + Send + Sync + Clone + 'static,
 {
-    pub(crate) fn cache_for_address(&self, address: Address) -> PrecompileCache<S> {
+    /// Get the precompile cache for the given address.
+    pub fn cache_for_address(&self, address: Address) -> PrecompileCache<S> {
         // Try just using `.get` first to avoid acquiring a write lock.
         if let Some(cache) = self.0.get(&address) {
             return cache.clone();
@@ -90,7 +91,7 @@ impl<S> CacheEntry<S> {
 
 /// A cache for precompile inputs / outputs.
 #[derive(Debug)]
-pub(crate) struct CachedPrecompile<S>
+pub struct CachedPrecompile<S>
 where
     S: Eq + Hash + std::fmt::Debug + Send + Sync + Clone + 'static,
 {
@@ -109,7 +110,7 @@ where
     S: Eq + Hash + std::fmt::Debug + Send + Sync + Clone + 'static,
 {
     /// `CachedPrecompile` constructor.
-    pub(crate) const fn new(
+    pub const fn new(
         precompile: DynPrecompile,
         cache: PrecompileCache<S>,
         spec_id: S,
@@ -118,7 +119,8 @@ where
         Self { precompile, cache, spec_id, metrics }
     }
 
-    pub(crate) fn wrap(
+    /// Wrap the given precompile in a cached precompile.
+    pub fn wrap(
         precompile: DynPrecompile,
         cache: PrecompileCache<S>,
         spec_id: S,
@@ -196,18 +198,18 @@ where
 /// Metrics for the cached precompile.
 #[derive(reth_metrics::Metrics, Clone)]
 #[metrics(scope = "sync.caching")]
-pub(crate) struct CachedPrecompileMetrics {
+pub struct CachedPrecompileMetrics {
     /// Precompile cache hits
-    precompile_cache_hits: metrics::Counter,
+    pub precompile_cache_hits: metrics::Counter,
 
     /// Precompile cache misses
-    precompile_cache_misses: metrics::Counter,
+    pub precompile_cache_misses: metrics::Counter,
 
     /// Precompile cache size. Uses the LRU cache length as the size metric.
-    precompile_cache_size: metrics::Gauge,
+    pub precompile_cache_size: metrics::Gauge,
 
     /// Precompile execution errors.
-    precompile_errors: metrics::Counter,
+    pub precompile_errors: metrics::Counter,
 }
 
 impl CachedPrecompileMetrics {
@@ -215,7 +217,7 @@ impl CachedPrecompileMetrics {
     ///
     /// Adds address as an `address` label padded with zeros to at least two hex symbols, prefixed
     /// by `0x`.
-    pub(crate) fn new_with_address(address: Address) -> Self {
+    pub fn new_with_address(address: Address) -> Self {
         Self::new_with_labels(&[("address", format!("0x{address:02x}"))])
     }
 }

crates/engine/tree/src/tree/state.rs

@@ -6,7 +6,7 @@ use alloy_primitives::{
     map::{HashMap, HashSet},
     BlockNumber, B256,
 };
-use reth_chain_state::{EthPrimitives, ExecutedBlock};
+use reth_chain_state::{DeferredTrieData, EthPrimitives, ExecutedBlock, LazyOverlay};
 use reth_primitives_traits::{AlloyBlockHeader, NodePrimitives, SealedHeader};
 use std::{
     collections::{btree_map, hash_map, BTreeMap, VecDeque},
@@ -38,40 +38,44 @@ pub struct TreeState<N: NodePrimitives = EthPrimitives> {
     pub(crate) current_canonical_head: BlockNumHash,
     /// The engine API variant of this handler
     pub(crate) engine_kind: EngineApiKind,
+    /// Pre-computed lazy overlay for the canonical head.
+    ///
+    /// This is optimistically prepared after the canonical head changes, so that
+    /// the next payload building on the canonical head can use it immediately
+    /// without recomputing.
+    pub(crate) cached_canonical_overlay: Option<PreparedCanonicalOverlay>,
 }
 
 impl<N: NodePrimitives> TreeState<N> {
     /// Returns a new, empty tree state that points to the given canonical head.
-    pub(crate) fn new(current_canonical_head: BlockNumHash, engine_kind: EngineApiKind) -> Self {
+    pub fn new(current_canonical_head: BlockNumHash, engine_kind: EngineApiKind) -> Self {
         Self {
             blocks_by_hash: HashMap::default(),
             blocks_by_number: BTreeMap::new(),
             current_canonical_head,
             parent_to_child: HashMap::default(),
             engine_kind,
+            cached_canonical_overlay: None,
         }
     }
 
     /// Resets the state and points to the given canonical head.
-    pub(crate) fn reset(&mut self, current_canonical_head: BlockNumHash) {
+    pub fn reset(&mut self, current_canonical_head: BlockNumHash) {
         *self = Self::new(current_canonical_head, self.engine_kind);
     }
 
     /// Returns the number of executed blocks stored.
-    pub(crate) fn block_count(&self) -> usize {
+    pub fn block_count(&self) -> usize {
         self.blocks_by_hash.len()
     }
 
     /// Returns the [`ExecutedBlock`] by hash.
-    pub(crate) fn executed_block_by_hash(&self, hash: B256) -> Option<&ExecutedBlock<N>> {
+    pub fn executed_block_by_hash(&self, hash: B256) -> Option<&ExecutedBlock<N>> {
         self.blocks_by_hash.get(&hash)
     }
 
     /// Returns the sealed block header by hash.
-    pub(crate) fn sealed_header_by_hash(
-        &self,
-        hash: &B256,
-    ) -> Option<SealedHeader<N::BlockHeader>> {
+    pub fn sealed_header_by_hash(&self, hash: &B256) -> Option<SealedHeader<N::BlockHeader>> {
         self.blocks_by_hash.get(hash).map(|b| b.sealed_block().sealed_header().clone())
     }
 
@@ -80,7 +84,7 @@ impl<N: NodePrimitives> TreeState<N> {
     /// highest persisted block connected to this chain.
     ///
     /// Returns `None` if the block for the given hash is not found.
-    pub(crate) fn blocks_by_hash(&self, hash: B256) -> Option<(B256, Vec<ExecutedBlock<N>>)> {
+    pub fn blocks_by_hash(&self, hash: B256) -> Option<(B256, Vec<ExecutedBlock<N>>)> {
         let block = self.blocks_by_hash.get(&hash).cloned()?;
         let mut parent_hash = block.recovered_block().parent_hash();
         let mut blocks = vec![block];
@@ -92,8 +96,68 @@ impl<N: NodePrimitives> TreeState<N> {
         Some((parent_hash, blocks))
     }
 
+    /// Prepares a cached lazy overlay for the current canonical head.
+    ///
+    /// This should be called after the canonical head changes to optimistically
+    /// prepare the overlay for the next payload that will likely build on it.
+    ///
+    /// Returns a clone of the [`LazyOverlay`] so the caller can spawn a background
+    /// task to trigger computation via [`LazyOverlay::get`]. This ensures the overlay
+    /// is actually computed before the next payload arrives.
+    pub(crate) fn prepare_canonical_overlay(&mut self) -> Option<LazyOverlay> {
+        let canonical_hash = self.current_canonical_head.hash;
+
+        // Get blocks leading to the canonical head
+        let Some((anchor_hash, blocks)) = self.blocks_by_hash(canonical_hash) else {
+            // Canonical head not in memory (persisted), no overlay needed
+            self.cached_canonical_overlay = None;
+            return None;
+        };
+
+        // Extract deferred trie data handles from blocks (newest to oldest)
+        let handles: Vec<DeferredTrieData> = blocks.iter().map(|b| b.trie_data_handle()).collect();
+
+        let overlay = LazyOverlay::new(anchor_hash, handles);
+        self.cached_canonical_overlay = Some(PreparedCanonicalOverlay {
+            parent_hash: canonical_hash,
+            overlay: overlay.clone(),
+            anchor_hash,
+        });
+
+        debug!(
+            target: "engine::tree",
+            %canonical_hash,
+            %anchor_hash,
+            num_blocks = blocks.len(),
+            "Prepared cached canonical overlay"
+        );
+
+        Some(overlay)
+    }
+
+    /// Returns the cached overlay if it matches the requested parent hash and anchor.
+    ///
+    /// Both parent hash and anchor hash must match to ensure the overlay is valid.
+    /// This prevents using a stale overlay after persistence has advanced the anchor.
+    pub(crate) fn get_cached_overlay(
+        &self,
+        parent_hash: B256,
+        expected_anchor: B256,
+    ) -> Option<&PreparedCanonicalOverlay> {
+        self.cached_canonical_overlay.as_ref().filter(|cached| {
+            cached.parent_hash == parent_hash && cached.anchor_hash == expected_anchor
+        })
+    }
+
+    /// Invalidates the cached overlay.
+    ///
+    /// Should be called when the anchor changes (e.g., after persistence).
+    pub(crate) fn invalidate_cached_overlay(&mut self) {
+        self.cached_canonical_overlay = None;
+    }
+
     /// Insert executed block into the state.
-    pub(crate) fn insert_executed(&mut self, executed: ExecutedBlock<N>) {
+    pub fn insert_executed(&mut self, executed: ExecutedBlock<N>) {
         let hash = executed.recovered_block().hash();
         let parent_hash = executed.recovered_block().parent_hash();
         let block_number = executed.recovered_block().number();
@@ -149,7 +213,7 @@ impl<N: NodePrimitives> TreeState<N> {
     }
 
     /// Returns whether or not the hash is part of the canonical chain.
-    pub(crate) fn is_canonical(&self, hash: B256) -> bool {
+    pub fn is_canonical(&self, hash: B256) -> bool {
         let mut current_block = self.current_canonical_head.hash;
         if current_block == hash {
             return true
@@ -167,11 +231,7 @@ impl<N: NodePrimitives> TreeState<N> {
 
     /// Removes canonical blocks below the upper bound, only if the last persisted hash is
     /// part of the canonical chain.
-    pub(crate) fn remove_canonical_until(
-        &mut self,
-        upper_bound: BlockNumber,
-        last_persisted_hash: B256,
-    ) {
+    pub fn remove_canonical_until(&mut self, upper_bound: BlockNumber, last_persisted_hash: B256) {
         debug!(target: "engine::tree", ?upper_bound, ?last_persisted_hash, "Removing canonical blocks from the tree");
 
         // If the last persisted hash is not canonical, then we don't want to remove any canonical
@@ -196,7 +256,7 @@ impl<N: NodePrimitives> TreeState<N> {
 
     /// Removes all blocks that are below the finalized block, as well as removing non-canonical
     /// sidechains that fork from below the finalized block.
-    pub(crate) fn prune_finalized_sidechains(&mut self, finalized_num_hash: BlockNumHash) {
+    pub fn prune_finalized_sidechains(&mut self, finalized_num_hash: BlockNumHash) {
         let BlockNumHash { number: finalized_num, hash: finalized_hash } = finalized_num_hash;
 
         // We remove disconnected sidechains in three steps:
@@ -256,7 +316,7 @@ impl<N: NodePrimitives> TreeState<N> {
     /// NOTE: if the finalized block is greater than the upper bound, the only blocks that will be
     /// removed are canonical blocks and sidechains that fork below the `upper_bound`. This is the
     /// same behavior as if the `finalized_num` were `Some(upper_bound)`.
-    pub(crate) fn remove_until(
+    pub fn remove_until(
         &mut self,
         upper_bound: BlockNumHash,
         last_persisted_hash: B256,
@@ -288,25 +348,28 @@ impl<N: NodePrimitives> TreeState<N> {
         if let Some(finalized_num_hash) = finalized_num_hash {
             self.prune_finalized_sidechains(finalized_num_hash);
         }
+
+        // Invalidate the cached overlay since blocks were removed and the anchor may have changed
+        self.invalidate_cached_overlay();
     }
 
     /// Updates the canonical head to the given block.
-    pub(crate) const fn set_canonical_head(&mut self, new_head: BlockNumHash) {
+    pub const fn set_canonical_head(&mut self, new_head: BlockNumHash) {
         self.current_canonical_head = new_head;
     }
 
     /// Returns the tracked canonical head.
-    pub(crate) const fn canonical_head(&self) -> &BlockNumHash {
+    pub const fn canonical_head(&self) -> &BlockNumHash {
         &self.current_canonical_head
     }
 
     /// Returns the block hash of the canonical head.
-    pub(crate) const fn canonical_block_hash(&self) -> B256 {
+    pub const fn canonical_block_hash(&self) -> B256 {
         self.canonical_head().hash
     }
 
     /// Returns the block number of the canonical head.
-    pub(crate) const fn canonical_block_number(&self) -> BlockNumber {
+    pub const fn canonical_block_number(&self) -> BlockNumber {
         self.canonical_head().number
     }
 }
@@ -316,7 +379,7 @@ impl<N: NodePrimitives> TreeState<N> {
     /// Determines if the second block is a descendant of the first block.
     ///
     /// If the two blocks are the same, this returns `false`.
-    pub(crate) fn is_descendant(
+    pub fn is_descendant(
         &self,
         first: BlockNumHash,
         second: alloy_eips::eip1898::BlockWithParent,
@@ -355,6 +418,39 @@ impl<N: NodePrimitives> TreeState<N> {
     }
 }
 
+/// Pre-computed lazy overlay for the canonical head block.
+///
+/// This is prepared **optimistically** when the canonical head changes, allowing
+/// the next payload (which typically builds on the canonical head) to reuse
+/// the pre-computed overlay immediately without re-traversing in-memory blocks.
+///
+/// The overlay captures deferred trie data handles from all in-memory blocks
+/// between the canonical head and the persisted anchor. When a new payload
+/// arrives building on the canonical head, this cached overlay can be used
+/// directly instead of calling `blocks_by_hash` and collecting handles again.
+///
+/// # Invalidation
+///
+/// The cached overlay is invalidated when:
+/// - Persistence completes (anchor changes)
+/// - The canonical head changes to a different block
+#[derive(Debug, Clone)]
+pub struct PreparedCanonicalOverlay {
+    /// The block hash for which this overlay is prepared as a parent.
+    ///
+    /// When a payload arrives with this parent hash, the overlay can be reused.
+    pub parent_hash: B256,
+    /// The pre-computed lazy overlay containing deferred trie data handles.
+    ///
+    /// This is computed optimistically after `set_canonical_head` so subsequent
+    /// payloads don't need to re-collect the handles.
+    pub overlay: LazyOverlay,
+    /// The anchor hash (persisted ancestor) this overlay is based on.
+    ///
+    /// Used to verify the overlay is still valid (anchor hasn't changed due to persistence).
+    pub anchor_hash: B256,
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

crates/engine/tree/src/tree/tests.rs

@@ -259,6 +259,7 @@ impl TestHarness {
             current_canonical_head: blocks.last().unwrap().recovered_block().num_hash(),
             parent_to_child,
             engine_kind: EngineApiKind::Ethereum,
+            cached_canonical_overlay: None,
         };
 
         let last_executed_block = blocks.last().unwrap().clone();

crates/engine/tree/src/tree/trie_updates.rs

@@ -98,7 +98,7 @@ impl StorageTrieUpdatesDiff {
 
 /// Compares the trie updates from state root task, regular state root calculation and database,
 /// and logs the differences if there's any.
-pub(super) fn compare_trie_updates(
+pub(crate) fn compare_trie_updates(
     trie_cursor_factory: impl TrieCursorFactory,
     task: TrieUpdates,
     regular: TrieUpdates,
@@ -186,7 +186,8 @@ fn compare_storage_trie_updates<C: TrieCursor>(
     task: &mut StorageTrieUpdates,
     regular: &mut StorageTrieUpdates,
 ) -> Result<StorageTrieUpdatesDiff, DatabaseError> {
-    let database_not_exists = trie_cursor()?.next()?.is_none();
+    // Check if the storage trie exists by seeking to the first entry
+    let database_not_exists = trie_cursor()?.seek(Nibbles::default())?.is_none();
     let mut diff = StorageTrieUpdatesDiff {
         // If the deletion is a no-op, meaning that the entry is not in the
         // database, do not add it to the diff.

crates/era-downloader/Cargo.toml

@@ -20,8 +20,6 @@ reth-era.workspace = true
 # http
 bytes.workspace = true
 reqwest.workspace = true
-reqwest.default-features = false
-reqwest.features = ["stream", "rustls-tls-native-roots"]
 
 # async
 tokio.workspace = true

crates/ethereum/cli/src/app.rs

@@ -86,7 +86,7 @@ where
         mut self,
         components: impl CliComponentsBuilder<N>,
         launcher: impl AsyncFnOnce(
-            WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+            WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
             Ext,
         ) -> Result<()>,
     ) -> Result<()>
@@ -132,7 +132,7 @@ pub(crate) fn run_commands_with<C, Ext, Rpc, N, SubCmd>(
     runner: CliRunner,
     components: impl CliComponentsBuilder<N>,
     launcher: impl AsyncFnOnce(
-        WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+        WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
         Ext,
     ) -> Result<()>,
 ) -> Result<()>
@@ -174,7 +174,7 @@ where
         }
         Commands::P2P(command) => runner.run_until_ctrl_c(command.execute::<N>()),
         Commands::Config(command) => runner.run_until_ctrl_c(command.execute()),
-        Commands::Prune(command) => runner.run_until_ctrl_c(command.execute::<N>()),
+        Commands::Prune(command) => runner.run_command_until_exit(|ctx| command.execute::<N>(ctx)),
         #[cfg(feature = "dev")]
         Commands::TestVectors(command) => runner.run_until_ctrl_c(command.execute()),
         Commands::ReExecute(command) => runner.run_until_ctrl_c(command.execute::<N>(components)),

crates/ethereum/cli/src/interface.rs

@@ -131,7 +131,7 @@ impl<
     /// ````
     pub fn run<L, Fut>(self, launcher: L) -> eyre::Result<()>
     where
-        L: FnOnce(WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>, Ext) -> Fut,
+        L: FnOnce(WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>, Ext) -> Fut,
         Fut: Future<Output = eyre::Result<()>>,
         C: ChainSpecParser<ChainSpec = ChainSpec>,
     {
@@ -148,7 +148,7 @@ impl<
         self,
         components: impl CliComponentsBuilder<N>,
         launcher: impl AsyncFnOnce(
-            WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+            WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
             Ext,
         ) -> eyre::Result<()>,
     ) -> eyre::Result<()>
@@ -180,7 +180,7 @@ impl<
     /// ```
     pub fn with_runner<L, Fut>(self, runner: CliRunner, launcher: L) -> eyre::Result<()>
     where
-        L: FnOnce(WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>, Ext) -> Fut,
+        L: FnOnce(WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>, Ext) -> Fut,
         Fut: Future<Output = eyre::Result<()>>,
         C: ChainSpecParser<ChainSpec = ChainSpec>,
     {
@@ -196,7 +196,7 @@ impl<
         runner: CliRunner,
         components: impl CliComponentsBuilder<N>,
         launcher: impl AsyncFnOnce(
-            WithLaunchContext<NodeBuilder<Arc<DatabaseEnv>, C::ChainSpec>>,
+            WithLaunchContext<NodeBuilder<DatabaseEnv, C::ChainSpec>>,
             Ext,
         ) -> eyre::Result<()>,
     ) -> eyre::Result<()>

crates/ethereum/node/src/node.rs

@@ -119,10 +119,9 @@ impl EthereumNode {
     /// use reth_db::open_db_read_only;
     /// use reth_node_ethereum::EthereumNode;
     /// use reth_provider::providers::{RocksDBProvider, StaticFileProvider};
-    /// use std::sync::Arc;
     ///
     /// let factory = EthereumNode::provider_factory_builder()
-    ///     .db(Arc::new(open_db_read_only("db", Default::default()).unwrap()))
+    ///     .db(open_db_read_only("db", Default::default()).unwrap())
     ///     .chainspec(ChainSpecBuilder::mainnet().build().into())
     ///     .static_file(StaticFileProvider::read_only("db/static_files", false).unwrap())
     ///     .rocksdb_provider(RocksDBProvider::builder("db/rocksdb").build().unwrap())

crates/ethereum/node/tests/e2e/blobs.rs

@@ -100,10 +100,12 @@ async fn can_send_legacy_sidecar_post_activation() -> eyre::Result<()> {
         ChainSpecBuilder::default().chain(MAINNET.chain).genesis(genesis).osaka_activated().build(),
     );
     let genesis_hash = chain_spec.genesis_hash();
-    let node_config = NodeConfig::test()
-        .with_chain(chain_spec)
-        .with_unused_ports()
-        .with_rpc(RpcServerArgs::default().with_unused_ports().with_http());
+    let node_config = NodeConfig::test().with_chain(chain_spec).with_unused_ports().with_rpc(
+        RpcServerArgs::default()
+            .with_unused_ports()
+            .with_http()
+            .with_force_blob_sidecar_upcasting(),
+    );
     let NodeHandle { node, node_exit_future: _ } = NodeBuilder::new(node_config.clone())
         .testing_node(exec.clone())
         .node(EthereumNode::default())
@@ -125,7 +127,7 @@ async fn can_send_legacy_sidecar_post_activation() -> eyre::Result<()> {
     let blob_tx_hash = node.rpc.inject_tx(blob_tx).await?;
     // fetch it from rpc
     let envelope = node.rpc.envelope_by_hash(blob_tx_hash).await?;
-    // assert that sidecar was converted to eip7594
+    // assert that sidecar was converted to eip7594 (force upcasting is enabled)
     assert!(envelope.as_eip4844().unwrap().tx().sidecar().unwrap().is_eip7594());
     // validate sidecar
     TransactionTestContext::validate_sidecar(envelope);
@@ -161,10 +163,12 @@ async fn blob_conversion_at_osaka() -> eyre::Result<()> {
             .build(),
     );
     let genesis_hash = chain_spec.genesis_hash();
-    let node_config = NodeConfig::test()
-        .with_chain(chain_spec)
-        .with_unused_ports()
-        .with_rpc(RpcServerArgs::default().with_unused_ports().with_http());
+    let node_config = NodeConfig::test().with_chain(chain_spec).with_unused_ports().with_rpc(
+        RpcServerArgs::default()
+            .with_unused_ports()
+            .with_http()
+            .with_force_blob_sidecar_upcasting(),
+    );
     let NodeHandle { node, node_exit_future: _ } = NodeBuilder::new(node_config.clone())
         .testing_node(exec.clone())
         .node(EthereumNode::default())

crates/ethereum/node/tests/e2e/eth.rs

@@ -1,4 +1,4 @@
-use crate::utils::eth_payload_attributes;
+use crate::utils::{advance_with_random_transactions, eth_payload_attributes};
 use alloy_eips::eip7685::RequestsOrHash;
 use alloy_genesis::Genesis;
 use alloy_primitives::{Address, B256};
@@ -6,8 +6,9 @@ use alloy_rpc_types_engine::{PayloadAttributes, PayloadStatusEnum};
 use jsonrpsee_core::client::ClientT;
 use reth_chainspec::{ChainSpecBuilder, EthChainSpec, MAINNET};
 use reth_e2e_test_utils::{
-    node::NodeTestContext, setup, transaction::TransactionTestContext, wallet::Wallet,
+    node::NodeTestContext, setup, setup_engine, transaction::TransactionTestContext, wallet::Wallet,
 };
+use reth_node_api::TreeConfig;
 use reth_node_builder::{NodeBuilder, NodeHandle};
 use reth_node_core::{args::RpcServerArgs, node_config::NodeConfig};
 use reth_node_ethereum::EthereumNode;
@@ -256,3 +257,56 @@ async fn test_testing_build_block_v1_osaka() -> eyre::Result<()> {
 
     Ok(())
 }
+
+/// Tests that sparse trie allocation reuse works correctly across consecutive blocks.
+///
+/// This test exercises the sparse trie allocation reuse path by:
+/// 1. Starting a node with parallel state root computation enabled
+/// 2. Advancing multiple consecutive blocks with random transactions
+/// 3. Verifying that all blocks are successfully validated (state roots match)
+///
+/// Note: Trie structure reuse is currently disabled due to pruning creating blinded
+/// nodes. The preserved trie's allocations are still reused to reduce memory overhead,
+/// but the trie is cleared between blocks.
+#[tokio::test]
+async fn test_sparse_trie_reuse_across_blocks() -> eyre::Result<()> {
+    reth_tracing::init_test_tracing();
+
+    // Use parallel state root (non-legacy) with pruning enabled
+    let tree_config = TreeConfig::default()
+        .with_legacy_state_root(false)
+        .with_sparse_trie_prune_depth(2)
+        .with_sparse_trie_max_storage_tries(100);
+
+    let (mut nodes, _tasks, _wallet) = setup_engine::<EthereumNode>(
+        1,
+        Arc::new(
+            ChainSpecBuilder::default()
+                .chain(MAINNET.chain)
+                .genesis(serde_json::from_str(include_str!("../assets/genesis.json")).unwrap())
+                .cancun_activated()
+                .build(),
+        ),
+        false,
+        tree_config,
+        eth_payload_attributes,
+    )
+    .await?;
+
+    let mut node = nodes.pop().unwrap();
+
+    // Use a seeded RNG for reproducibility
+    let mut rng = rand::rng();
+
+    // Advance multiple consecutive blocks with random transactions.
+    // This exercises the sparse trie reuse path where each block's pruned trie
+    // is reused for the next block's state root computation.
+    let num_blocks = 5;
+    advance_with_random_transactions(&mut node, num_blocks, &mut rng, true).await?;
+
+    // Verify the chain advanced correctly
+    let best_block = node.inner.provider.best_block_number()?;
+    assert_eq!(best_block, num_blocks as u64, "Expected {} blocks, got {}", num_blocks, best_block);
+
+    Ok(())
+}

crates/ethereum/primitives/src/receipt.rs

@@ -511,9 +511,8 @@ mod compact {
             total_length += flags.len() + buffer.len();
             buf.put_slice(&flags);
             if zstd {
-                reth_zstd_compressors::RECEIPT_COMPRESSOR.with(|compressor| {
-                    let compressed =
-                        compressor.borrow_mut().compress(&buffer).expect("Failed to compress.");
+                reth_zstd_compressors::with_receipt_compressor(|compressor| {
+                    let compressed = compressor.compress(&buffer).expect("Failed to compress.");
                     buf.put(compressed.as_slice());
                 });
             } else {
@@ -525,8 +524,7 @@ mod compact {
         fn from_compact(buf: &[u8], _len: usize) -> (Self, &[u8]) {
             let (flags, mut buf) = ReceiptFlags::from(buf);
             if flags.__zstd() != 0 {
-                reth_zstd_compressors::RECEIPT_DECOMPRESSOR.with(|decompressor| {
-                    let decompressor = &mut decompressor.borrow_mut();
+                reth_zstd_compressors::with_receipt_decompressor(|decompressor| {
                     let decompressed = decompressor.decompress(buf);
                     let original_buf = buf;
                     let mut buf: &[u8] = decompressed;

crates/ethereum/primitives/src/transaction.rs

@@ -577,19 +577,11 @@ impl reth_codecs::Compact for TransactionSigned {
 
         let tx_bits = if zstd_bit {
             let mut tmp = Vec::with_capacity(256);
-            if cfg!(feature = "std") {
-                reth_zstd_compressors::TRANSACTION_COMPRESSOR.with(|compressor| {
-                    let mut compressor = compressor.borrow_mut();
-                    let tx_bits = self.transaction.to_compact(&mut tmp);
-                    buf.put_slice(&compressor.compress(&tmp).expect("Failed to compress"));
-                    tx_bits as u8
-                })
-            } else {
-                let mut compressor = reth_zstd_compressors::create_tx_compressor();
+            reth_zstd_compressors::with_tx_compressor(|compressor| {
                 let tx_bits = self.transaction.to_compact(&mut tmp);
                 buf.put_slice(&compressor.compress(&tmp).expect("Failed to compress"));
                 tx_bits as u8
-            }
+            })
         } else {
             self.transaction.to_compact(buf) as u8
         };
@@ -611,26 +603,13 @@ impl reth_codecs::Compact for TransactionSigned {
 
         let zstd_bit = bitflags >> 3;
         let (transaction, buf) = if zstd_bit != 0 {
-            if cfg!(feature = "std") {
-                reth_zstd_compressors::TRANSACTION_DECOMPRESSOR.with(|decompressor| {
-                    let mut decompressor = decompressor.borrow_mut();
-
-                    // TODO: enforce that zstd is only present at a "top" level type
-
-                    let transaction_type = (bitflags & 0b110) >> 1;
-                    let (transaction, _) =
-                        Transaction::from_compact(decompressor.decompress(buf), transaction_type);
-
-                    (transaction, buf)
-                })
-            } else {
-                let mut decompressor = reth_zstd_compressors::create_tx_decompressor();
+            reth_zstd_compressors::with_tx_decompressor(|decompressor| {
+                // TODO: enforce that zstd is only present at a "top" level type
                 let transaction_type = (bitflags & 0b110) >> 1;
                 let (transaction, _) =
                     Transaction::from_compact(decompressor.decompress(buf), transaction_type);
-
                 (transaction, buf)
-            }
+            })
         } else {
             let transaction_type = bitflags >> 1;
             Transaction::from_compact(buf, transaction_type)

crates/evm/execution-errors/src/trie.rs

@@ -171,7 +171,7 @@ pub enum SparseTrieErrorKind {
         /// Path to the node.
         path: Nibbles,
         /// Node that was at the path when revealing.
-        node: Box<dyn core::fmt::Debug + Send>,
+        node: Box<dyn core::fmt::Debug + Send + Sync>,
     },
     /// RLP error.
     #[error(transparent)]
@@ -184,7 +184,7 @@ pub enum SparseTrieErrorKind {
     },
     /// Other.
     #[error(transparent)]
-    Other(#[from] Box<dyn core::error::Error + Send>),
+    Other(#[from] Box<dyn core::error::Error + Send + Sync>),
 }
 
 /// Trie witness errors.

crates/exex/test-utils/src/lib.rs

@@ -251,7 +251,7 @@ pub async fn test_exex_context_with_chain_spec(
         db,
         chain_spec.clone(),
         StaticFileProvider::read_write(static_dir.keep()).expect("static file provider"),
-        RocksDBProvider::builder(rocksdb_dir.keep()).build().unwrap(),
+        RocksDBProvider::builder(rocksdb_dir.keep()).with_default_tables().build().unwrap(),
     )?;
 
     let genesis_hash = init_genesis(&provider_factory)?;

crates/net/downloaders/src/file_client.rs

@@ -83,6 +83,28 @@ impl From<&'static str> for FileClientError {
 }
 
 impl<B: FullBlock> FileClient<B> {
+    /// Create a new file client from a slice of sealed blocks.
+    pub fn from_blocks(blocks: impl IntoIterator<Item = SealedBlock<B>>) -> Self {
+        let blocks: Vec<_> = blocks.into_iter().collect();
+        let capacity = blocks.len();
+
+        let mut headers = HashMap::with_capacity(capacity);
+        let mut hash_to_number = HashMap::with_capacity(capacity);
+        let mut bodies = HashMap::with_capacity(capacity);
+
+        for block in blocks {
+            let number = block.number();
+            let hash = block.hash();
+            let (header, body) = block.split_sealed_header_body();
+
+            headers.insert(number, header.into_header());
+            hash_to_number.insert(hash, number);
+            bodies.insert(hash, body);
+        }
+
+        Self { headers, hash_to_number, bodies }
+    }
+
     /// Create a new file client from a file path.
     pub async fn new<P: AsRef<Path>>(
         path: P,

crates/net/eth-wire/src/errors/p2p.rs

@@ -68,7 +68,7 @@ pub enum P2PStreamError {
     SendBufferFull,
 
     /// Disconnected error.
-    #[error("disconnected")]
+    #[error("disconnected: {0}")]
     Disconnected(DisconnectReason),
 
     /// Unknown disconnect reason error.

crates/net/network/src/swarm.rs

@@ -20,7 +20,7 @@ use std::{
     sync::Arc,
     task::{Context, Poll},
 };
-use tracing::{debug, trace};
+use tracing::trace;
 
 #[cfg_attr(doc, aquamarine::aquamarine)]
 /// Contains the connectivity related state of the network.
@@ -259,7 +259,7 @@ impl<N: NetworkPrimitives> Swarm<N> {
                 if self.sessions.is_valid_fork_id(fork_id) {
                     self.state_mut().peers_mut().set_discovered_fork_id(peer_id, fork_id);
                 } else {
-                    debug!(target: "net", ?peer_id, remote_fork_id=?fork_id, our_fork_id=?self.sessions.fork_id(), "fork id mismatch, removing peer");
+                    trace!(target: "net", ?peer_id, remote_fork_id=?fork_id, our_fork_id=?self.sessions.fork_id(), "fork id mismatch, removing peer");
                     self.state_mut().peers_mut().remove_peer(peer_id);
                 }
             }

crates/node/builder/src/builder/mod.rs

@@ -251,6 +251,8 @@ impl<DB, ChainSpec: EthChainSpec> NodeBuilder<DB, ChainSpec> {
     }
 
     /// Creates a preconfigured node for testing purposes with a specific datadir.
+    ///
+    /// The entire `datadir` will be cleaned up when the node is dropped.
     #[cfg(feature = "test-utils")]
     pub fn testing_node_with_datadir(
         mut self,
@@ -268,7 +270,7 @@ impl<DB, ChainSpec: EthChainSpec> NodeBuilder<DB, ChainSpec> {
         let data_dir =
             path.unwrap_or_chain_default(self.config.chain.chain(), self.config.datadir.clone());
 
-        let db = reth_db::test_utils::create_test_rw_db_with_path(data_dir.db());
+        let db = reth_db::test_utils::create_test_rw_db_with_datadir(data_dir.data_dir());
 
         WithLaunchContext { builder: self.with_database(db), task_executor }
     }

crates/node/builder/src/launch/engine.rs

@@ -303,20 +303,8 @@ impl EngineNodeLauncher {
             // the CL
             loop {
                 tokio::select! {
-                    shutdown_req = &mut shutdown_rx => {
-                        if let Ok(req) = shutdown_req {
-                            debug!(target: "reth::cli", "received engine shutdown request");
-                            engine_service.orchestrator_mut().handler_mut().handler_mut().on_event(
-                                FromOrchestrator::Terminate { tx: req.done_tx }.into()
-                            );
-                        }
-                    }
-                    payload = built_payloads.select_next_some() => {
-                        if let Some(executed_block) = payload.executed_block() {
-                            debug!(target: "reth::cli", block=?executed_block.recovered_block.num_hash(),  "inserting built payload");
-                            engine_service.orchestrator_mut().handler_mut().handler_mut().on_event(EngineApiRequest::InsertExecutedBlock(executed_block.into_executed_payload()).into());
-                        }
-                    }
+                    biased;
+
                     event = engine_service.next() => {
                         let Some(event) = event else { break };
                         debug!(target: "reth::cli", "Event: {event}");
@@ -364,6 +352,20 @@ impl EngineNodeLauncher {
                             }
                         }
                     }
+                    payload = built_payloads.select_next_some() => {
+                        if let Some(executed_block) = payload.executed_block() {
+                            debug!(target: "reth::cli", block=?executed_block.recovered_block.num_hash(),  "inserting built payload");
+                            engine_service.orchestrator_mut().handler_mut().handler_mut().on_event(EngineApiRequest::InsertExecutedBlock(executed_block.into_executed_payload()).into());
+                        }
+                    }
+                    shutdown_req = &mut shutdown_rx => {
+                        if let Ok(req) = shutdown_req {
+                            debug!(target: "reth::cli", "received engine shutdown request");
+                            engine_service.orchestrator_mut().handler_mut().handler_mut().on_event(
+                                FromOrchestrator::Terminate { tx: req.done_tx }.into()
+                            );
+                        }
+                    }
                 }
             }
 

crates/node/builder/src/node.rs

@@ -218,9 +218,9 @@ impl<Node: FullNodeComponents, AddOns: NodeAddOns<Node>> DerefMut for FullNode<N
 }
 
 /// Helper type alias to define [`FullNode`] for a given [`Node`].
-pub type FullNodeFor<N, DB = Arc<DatabaseEnv>> =
+pub type FullNodeFor<N, DB = DatabaseEnv> =
     FullNode<NodeAdapter<RethFullAdapter<DB, N>>, <N as Node<RethFullAdapter<DB, N>>>::AddOns>;
 
 /// Helper type alias to define [`NodeHandle`] for a given [`Node`].
-pub type NodeHandleFor<N, DB = Arc<DatabaseEnv>> =
+pub type NodeHandleFor<N, DB = DatabaseEnv> =
     NodeHandle<NodeAdapter<RethFullAdapter<DB, N>>, <N as Node<RethFullAdapter<DB, N>>>::AddOns>;

crates/node/builder/src/rpc.rs

@@ -1192,6 +1192,7 @@ impl<'a, N: FullNodeComponents<Types: NodeTypes<ChainSpec: Hardforks + EthereumH
             .pending_block_kind(self.config.pending_block_kind)
             .raw_tx_forwarder(self.config.raw_tx_forwarder)
             .evm_memory_limit(self.config.rpc_evm_memory_limit)
+            .force_blob_sidecar_upcasting(self.config.force_blob_sidecar_upcasting)
     }
 }
 

crates/node/core/src/args/engine.rs

@@ -1,7 +1,10 @@
 //! clap [Args](clap::Args) for engine purposes
 
 use clap::{builder::Resettable, Args};
-use reth_engine_primitives::{TreeConfig, DEFAULT_MULTIPROOF_TASK_CHUNK_SIZE};
+use reth_engine_primitives::{
+    TreeConfig, DEFAULT_MULTIPROOF_TASK_CHUNK_SIZE, DEFAULT_SPARSE_TRIE_MAX_STORAGE_TRIES,
+    DEFAULT_SPARSE_TRIE_PRUNE_DEPTH,
+};
 use std::sync::OnceLock;
 
 use crate::node_config::{
@@ -35,8 +38,11 @@ pub struct DefaultEngineValues {
     allow_unwind_canonical_header: bool,
     storage_worker_count: Option<usize>,
     account_worker_count: Option<usize>,
-    enable_proof_v2: bool,
+    disable_proof_v2: bool,
     cache_metrics_disabled: bool,
+    enable_sparse_trie_as_cache: bool,
+    sparse_trie_prune_depth: usize,
+    sparse_trie_max_storage_tries: usize,
 }
 
 impl DefaultEngineValues {
@@ -161,9 +167,9 @@ impl DefaultEngineValues {
         self
     }
 
-    /// Set whether to enable proof V2 by default
-    pub const fn with_enable_proof_v2(mut self, v: bool) -> Self {
-        self.enable_proof_v2 = v;
+    /// Set whether to disable proof V2 by default
+    pub const fn with_disable_proof_v2(mut self, v: bool) -> Self {
+        self.disable_proof_v2 = v;
         self
     }
 
@@ -172,6 +178,24 @@ impl DefaultEngineValues {
         self.cache_metrics_disabled = v;
         self
     }
+
+    /// Set whether to enable sparse trie as cache by default
+    pub const fn with_enable_sparse_trie_as_cache(mut self, v: bool) -> Self {
+        self.enable_sparse_trie_as_cache = v;
+        self
+    }
+
+    /// Set the sparse trie prune depth by default
+    pub const fn with_sparse_trie_prune_depth(mut self, v: usize) -> Self {
+        self.sparse_trie_prune_depth = v;
+        self
+    }
+
+    /// Set the maximum number of storage tries to retain after sparse trie pruning by default
+    pub const fn with_sparse_trie_max_storage_tries(mut self, v: usize) -> Self {
+        self.sparse_trie_max_storage_tries = v;
+        self
+    }
 }
 
 impl Default for DefaultEngineValues {
@@ -195,8 +219,11 @@ impl Default for DefaultEngineValues {
             allow_unwind_canonical_header: false,
             storage_worker_count: None,
             account_worker_count: None,
-            enable_proof_v2: false,
+            disable_proof_v2: false,
             cache_metrics_disabled: false,
+            enable_sparse_trie_as_cache: false,
+            sparse_trie_prune_depth: DEFAULT_SPARSE_TRIE_PRUNE_DEPTH,
+            sparse_trie_max_storage_tries: DEFAULT_SPARSE_TRIE_MAX_STORAGE_TRIES,
         }
     }
 }
@@ -317,13 +344,25 @@ pub struct EngineArgs {
     #[arg(long = "engine.account-worker-count", default_value = Resettable::from(DefaultEngineValues::get_global().account_worker_count.map(|v| v.to_string().into())))]
     pub account_worker_count: Option<usize>,
 
-    /// Enable V2 storage proofs for state root calculations
-    #[arg(long = "engine.enable-proof-v2", default_value_t = DefaultEngineValues::get_global().enable_proof_v2)]
-    pub enable_proof_v2: bool,
+    /// Disable V2 storage proofs for state root calculations
+    #[arg(long = "engine.disable-proof-v2", default_value_t = DefaultEngineValues::get_global().disable_proof_v2)]
+    pub disable_proof_v2: bool,
 
     /// Disable cache metrics recording, which can take up to 50ms with large cached state.
     #[arg(long = "engine.disable-cache-metrics", default_value_t = DefaultEngineValues::get_global().cache_metrics_disabled)]
     pub cache_metrics_disabled: bool,
+
+    /// Enable sparse trie as cache.
+    #[arg(long = "engine.enable-sparse-trie-as-cache", default_value_t = DefaultEngineValues::get_global().enable_sparse_trie_as_cache, conflicts_with = "disable_proof_v2")]
+    pub enable_sparse_trie_as_cache: bool,
+
+    /// Sparse trie prune depth.
+    #[arg(long = "engine.sparse-trie-prune-depth", default_value_t = DefaultEngineValues::get_global().sparse_trie_prune_depth, requires = "enable_sparse_trie_as_cache")]
+    pub sparse_trie_prune_depth: usize,
+
+    /// Maximum number of storage tries to retain after sparse trie pruning.
+    #[arg(long = "engine.sparse-trie-max-storage-tries", default_value_t = DefaultEngineValues::get_global().sparse_trie_max_storage_tries, requires = "enable_sparse_trie_as_cache")]
+    pub sparse_trie_max_storage_tries: usize,
 }
 
 #[allow(deprecated)]
@@ -348,8 +387,11 @@ impl Default for EngineArgs {
             allow_unwind_canonical_header,
             storage_worker_count,
             account_worker_count,
-            enable_proof_v2,
+            disable_proof_v2,
             cache_metrics_disabled,
+            enable_sparse_trie_as_cache,
+            sparse_trie_prune_depth,
+            sparse_trie_max_storage_tries,
         } = DefaultEngineValues::get_global().clone();
         Self {
             persistence_threshold,
@@ -374,8 +416,11 @@ impl Default for EngineArgs {
             allow_unwind_canonical_header,
             storage_worker_count,
             account_worker_count,
-            enable_proof_v2,
+            disable_proof_v2,
             cache_metrics_disabled,
+            enable_sparse_trie_as_cache,
+            sparse_trie_prune_depth,
+            sparse_trie_max_storage_tries,
         }
     }
 }
@@ -383,7 +428,7 @@ impl Default for EngineArgs {
 impl EngineArgs {
     /// Creates a [`TreeConfig`] from the engine arguments.
     pub fn tree_config(&self) -> TreeConfig {
-        let mut config = TreeConfig::default()
+        TreeConfig::default()
             .with_persistence_threshold(self.persistence_threshold)
             .with_memory_block_buffer_target(self.memory_block_buffer_target)
             .with_legacy_state_root(self.legacy_state_root_task_enabled)
@@ -400,20 +445,14 @@ impl EngineArgs {
             .with_always_process_payload_attributes_on_canonical_head(
                 self.always_process_payload_attributes_on_canonical_head,
             )
-            .with_unwind_canonical_header(self.allow_unwind_canonical_header);
-
-        if let Some(count) = self.storage_worker_count {
-            config = config.with_storage_worker_count(count);
-        }
-
-        if let Some(count) = self.account_worker_count {
-            config = config.with_account_worker_count(count);
-        }
-
-        config = config.with_enable_proof_v2(self.enable_proof_v2);
-        config = config.without_cache_metrics(self.cache_metrics_disabled);
-
-        config
+            .with_unwind_canonical_header(self.allow_unwind_canonical_header)
+            .with_storage_worker_count_opt(self.storage_worker_count)
+            .with_account_worker_count_opt(self.account_worker_count)
+            .with_disable_proof_v2(self.disable_proof_v2)
+            .without_cache_metrics(self.cache_metrics_disabled)
+            .with_enable_sparse_trie_as_cache(self.enable_sparse_trie_as_cache)
+            .with_sparse_trie_prune_depth(self.sparse_trie_prune_depth)
+            .with_sparse_trie_max_storage_tries(self.sparse_trie_max_storage_tries)
     }
 }
 
@@ -462,8 +501,11 @@ mod tests {
             allow_unwind_canonical_header: true,
             storage_worker_count: Some(16),
             account_worker_count: Some(8),
-            enable_proof_v2: false,
+            disable_proof_v2: false,
             cache_metrics_disabled: true,
+            enable_sparse_trie_as_cache: true,
+            sparse_trie_prune_depth: 10,
+            sparse_trie_max_storage_tries: 100,
         };
 
         let parsed_args = CommandParser::<EngineArgs>::parse_from([
@@ -494,6 +536,11 @@ mod tests {
             "--engine.account-worker-count",
             "8",
             "--engine.disable-cache-metrics",
+            "--engine.enable-sparse-trie-as-cache",
+            "--engine.sparse-trie-prune-depth",
+            "10",
+            "--engine.sparse-trie-max-storage-tries",
+            "100",
         ])
         .args;
 

crates/node/core/src/args/pruning.rs

@@ -5,7 +5,9 @@ use alloy_primitives::{Address, BlockNumber};
 use clap::{builder::RangedU64ValueParser, Args};
 use reth_chainspec::EthereumHardforks;
 use reth_config::config::PruneConfig;
-use reth_prune_types::{PruneMode, PruneModes, ReceiptsLogPruneConfig, MINIMUM_PRUNING_DISTANCE};
+use reth_prune_types::{
+    PruneMode, PruneModes, ReceiptsLogPruneConfig, MINIMUM_DISTANCE, MINIMUM_UNWIND_SAFE_DISTANCE,
+};
 use std::{collections::BTreeMap, ops::Not, sync::OnceLock};
 
 /// Global static pruning defaults
@@ -68,9 +70,9 @@ impl Default for DefaultPruningValues {
             full_prune_modes: PruneModes {
                 sender_recovery: Some(PruneMode::Full),
                 transaction_lookup: None,
-                receipts: Some(PruneMode::Distance(MINIMUM_PRUNING_DISTANCE)),
-                account_history: Some(PruneMode::Distance(MINIMUM_PRUNING_DISTANCE)),
-                storage_history: Some(PruneMode::Distance(MINIMUM_PRUNING_DISTANCE)),
+                receipts: Some(PruneMode::Distance(MINIMUM_UNWIND_SAFE_DISTANCE)),
+                account_history: Some(PruneMode::Distance(MINIMUM_UNWIND_SAFE_DISTANCE)),
+                storage_history: Some(PruneMode::Distance(MINIMUM_UNWIND_SAFE_DISTANCE)),
                 // This field is ignored when full_bodies_history_use_pre_merge is true
                 bodies_history: None,
                 receipts_log_filter: Default::default(),
@@ -79,10 +81,10 @@ impl Default for DefaultPruningValues {
             minimal_prune_modes: PruneModes {
                 sender_recovery: Some(PruneMode::Full),
                 transaction_lookup: Some(PruneMode::Full),
-                receipts: Some(PruneMode::Full),
-                account_history: Some(PruneMode::Distance(MINIMUM_PRUNING_DISTANCE)),
-                storage_history: Some(PruneMode::Distance(MINIMUM_PRUNING_DISTANCE)),
-                bodies_history: Some(PruneMode::Distance(MINIMUM_PRUNING_DISTANCE)),
+                receipts: Some(PruneMode::Distance(MINIMUM_DISTANCE)),
+                account_history: Some(PruneMode::Distance(MINIMUM_UNWIND_SAFE_DISTANCE)),
+                storage_history: Some(PruneMode::Distance(MINIMUM_UNWIND_SAFE_DISTANCE)),
+                bodies_history: Some(PruneMode::Distance(MINIMUM_UNWIND_SAFE_DISTANCE)),
                 receipts_log_filter: Default::default(),
             },
         }
@@ -93,7 +95,8 @@ impl Default for DefaultPruningValues {
 #[derive(Debug, Clone, Args, PartialEq, Eq, Default)]
 #[command(next_help_heading = "Pruning")]
 pub struct PruningArgs {
-    /// Run full node. Only the most recent [`MINIMUM_PRUNING_DISTANCE`] block states are stored.
+    /// Run full node. Only the most recent [`MINIMUM_UNWIND_SAFE_DISTANCE`] block states are
+    /// stored.
     #[arg(long, default_value_t = false, conflicts_with = "minimal")]
     pub full: bool,