Merge branch 'main' into dan/static-file-split

Co-authored-by: Amp <amp@ampcode.com>

.changelog/big-mules-cook.md

@@ -0,0 +1,4 @@
+---
+---
+
+Added site-level meta description for SEO.

.changelog/bold-frogs-run.md

@@ -0,0 +1,5 @@
+---
+reth-transaction-pool: patch
+---
+
+Renamed and documented validation methods for clarity. `validate_one_no_state` and `validate_one_against_state` are now public methods `validate_stateless` and `validate_stateful` with improved documentation explaining their respective validation phases.

.changelog/brave-dogs-fly.md

@@ -0,0 +1,10 @@
+---
+reth-engine-primitives: patch
+reth-engine-tree: patch
+reth-node-core: patch
+reth-trie-parallel: minor
+---
+
+Removed legacy proof calculation system and V2-specific configuration flags.
+
+Removed the legacy (non-V2) proof calculation code paths, simplified multiproof task architecture by removing the dual-mode system, and cleaned up V2-specific CLI flags (`--engine.disable-proof-v2`, `--engine.disable-trie-cache`) that are no longer needed. The codebase now exclusively uses V2 proofs with the sparse trie cache.

.changelog/calm-clams-buzz.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: patch
+---
+
+Refactored sparse trie node state tracking to use RLP nodes instead of hashes. Replaced `Option<B256>` hash fields with `SparseNodeState` enum that tracks either dirty nodes or cached RLP nodes with optional database storage flags. Added debug assertions to validate leaf path lengths and improved pruning logic to use node paths directly instead of path-hash tuples.

.changelog/config.toml

@@ -11,7 +11,7 @@ format = "per-crate"
 # Fixed groups: all always share the same version
 # reth binaries share version
 [[fixed]]
-members = ["reth", "op-reth"]
+members = ["reth"]
 
 # Packages to ignore (internal/test-only crates)
 ignore = [

.changelog/cool-owls-rest.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: patch
+---
+
+Fixed a bug in `merge_subtrie_updates` where source insertions did not cancel destination removals (and vice versa), causing inconsistent trie updates accumulated across multiple `root()` calls without intermediate `take_updates()`. Added a test covering the cross-cancellation behavior.

.changelog/cool-suns-rest.md

@@ -0,0 +1,5 @@
+---
+reth-transaction-pool: minor
+---
+
+Added support for optional custom stateless and stateful validation hooks in `EthTransactionValidator` via `set_additional_stateless_validation` and `set_additional_stateful_validation` methods. Also implemented a manual `Debug` impl to handle the non-`Debug` function pointer fields.

.changelog/cute-bears-paint.md

@@ -0,0 +1,17 @@
+---
+reth: minor
+reth-cli-commands: minor
+reth-e2e-test-utils: minor
+reth-ethereum-cli: minor
+reth-node-core: minor
+reth-optimism-bin: minor
+reth-optimism-cli: minor
+reth-prune: patch
+reth-stages: patch
+reth-storage-api: minor
+reth-storage-db-api: minor
+reth-storage-db-common: patch
+reth-storage-provider: patch
+---
+
+Introduced `--storage.v2` flag to control storage mode defaults, replacing the `edge` feature flag with `rocksdb` feature. The new flag enables v2 storage settings (static files + RocksDB routing) while individual `--static-files.*` and `--rocksdb.*` flags can still override defaults. Updated feature gates from `edge` to `rocksdb` across all affected crates.

.changelog/dark-ants-write.md

@@ -0,0 +1,5 @@
+---
+reth-tasks: patch
+---
+
+Added panic handler to all rayon thread pools that logs panics via `tracing::error` instead of aborting the process.

.changelog/dry-ants-bow.md

@@ -0,0 +1,5 @@
+---
+reth: patch
+---
+
+Removed Windows platform support from the codebase, including the Windows cross-compilation Dockerfile, build targets in Cross.toml and Makefile, and Windows-specific options in the bug report template.

.changelog/dry-ducks-write.md

@@ -0,0 +1,5 @@
+---
+reth-network: minor
+---
+
+Added reason label to backed_off_peers metric. The metric now tracks backed off peers by reason (too_many_peers, graceful_close, connection_error) to improve observability.

.changelog/dry-koalas-wink.md

@@ -0,0 +1,5 @@
+---
+reth-network-types: patch
+---
+
+Increased default maximum concurrent outbound dials from 15 to 30.

.changelog/dry-lakes-sleep.md

@@ -0,0 +1,6 @@
+---
+reth-trie-common: minor
+reth-trie: minor
+---
+
+Added `contains_range` method to `PrefixSet` for checking if any key falls within a half-open range. Added prefix set support to `ProofCalculator` via `with_prefix_set`, enabling stale cached hash invalidation and branch collapse detection when keys are inserted or removed; propagated storage prefix sets through `SyncAccountValueEncoder`.

.changelog/dull-clams-pack.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: patch
+---
+
+Refactored arena trie internals by adding a `BranchChildIdx::sibling()` helper, deduplicating `Index`/`NodeArena` type aliases, and replacing `is_empty()` with a `drop_root()` method. Fixed a bug where `cursor.pop()` was called before checking if the leaf was the root node, which could cause incorrect dirty-state propagation.

.changelog/dull-koalas-play.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: patch
+---
+
+Added recording of `SetRoot` operation in `ParallelSparseTrie::set_root` when the `trie-debug` feature is enabled.

.changelog/dull-seals-laugh.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: minor
+---
+
+Fixed a bug in `ArenaParallelSparseTrie` where subtrie updates that would completely empty a subtrie were incorrectly dispatched to parallel workers instead of being processed inline, preventing correct branch collapse detection when blinded siblings are present. Refactored the `SparseTrie` test suite to accept a `fn() -> T` factory instead of requiring `T: Default`, enabling a new `arena_parallel_sparse_trie_always_parallel` test variant that exercises all tests with parallelism thresholds set to 1. Added `test_branch_collapse_multi_empty_subtries_blinded_remaining` to cover the case where removing multiple revealed leaves empties their subtries and leaves a single blinded sibling requiring a proof.

.changelog/eager-mules-fold.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse-parallel: patch
+---
+
+Fixed parallel sparse trie to skip revealing disconnected leaves by checking parent branch reachability before inserting leaf nodes.

.changelog/easy-clouds-meow.md

@@ -0,0 +1,5 @@
+---
+ef-tests: patch
+---
+
+Removed reth-stateless crate and stateless validation from ef-tests.

.changelog/evil-fish-smile.md

@@ -0,0 +1,6 @@
+---
+reth-rpc-convert: minor
+reth-storage-rpc-provider: minor
+---
+
+Replaced the separate `TryFromBlockResponse`, `TryFromReceiptResponse`, and `TryFromTransactionResponse` traits with a unified `RpcResponseConverter` trait and default `EthRpcConverter` implementation. Removed the `op-alloy-network` dependency and refactored `RpcBlockchainProvider` to store a dynamic converter instance instead of relying on per-type trait bounds.

.changelog/evil-pigs-cry.md

@@ -0,0 +1,6 @@
+---
+reth-engine-tree: patch
+reth-trie-sparse-parallel: patch
+---
+
+Added tracing spans and debug logs to sparse trie operations for better observability during parallel state root computation.

.changelog/fair-winds-growl.md

@@ -0,0 +1,6 @@
+---
+reth-exex: patch
+reth-exex-types: patch
+---
+
+Added configurable backfill thresholds to ExEx notifications stream and added regression tests for state provider parity between pipeline and backfill execution paths.

.changelog/fair-wolves-smile.md

@@ -0,0 +1,5 @@
+---
+reth-payload-builder: minor
+---
+
+Added observability metrics for payload resolve latency and new payload job creation latency to the payload builder service.

.changelog/fast-fish-cry.md

@@ -0,0 +1,4 @@
+---
+---
+
+Added WebSocket subscription integration tests for eth_subscribe.

.changelog/fast-seals-play.md

@@ -0,0 +1,5 @@
+---
+reth-transaction-pool: minor
+---
+
+Added `consensus_ref` method to `PoolTransaction` trait for borrowing consensus transactions without cloning.

.changelog/fast-waves-smile.md

@@ -0,0 +1,4 @@
+---
+---
+
+Improved nightly Docker build failure Slack notification with more detailed formatting and context.

.changelog/fine-horses-whisper.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: patch
+---
+
+Fixed another branch collapse edge case where `check_subtrie_collapse_needs_proof` incorrectly compared removal count against total update count (including `Touched` entries), causing it to skip proof requests for blinded siblings and panic when the subtrie emptied. Added a regression test covering the removals + `Touched` + blinded sibling scenario.

.changelog/fine-koalas-shout.md

@@ -0,0 +1,10 @@
+---
+reth-chain-state: minor
+reth-engine-primitives: minor
+reth-engine-tree: minor
+reth-node-core: minor
+reth-node-events: minor
+reth: patch
+---
+
+Added configurable slow block logging (`--engine.slow-block-threshold`) that emits a structured `warn!` log with detailed timing, state-operation counts, and cache hit-rate metrics for blocks whose total processing time exceeds the threshold. Introduced `ExecutionTimingStats`, `CacheStats`, `StateProviderStats`, and `SlowBlockInfo` types to carry execution statistics from block validation through persistence, and refactored `PersistenceResult` to carry commit duration alongside the last persisted block.

.changelog/fix-download-api-url.md

@@ -0,0 +1,5 @@
+---
+reth-cli-commands: patch
+---
+
+Added `snapshot_api_url` field to `DownloadDefaults` so downstream projects can override the snapshot discovery API endpoint. Previously, `discover_manifest_url`, `fetch_snapshot_api_entries`, and `print_snapshot_listing` used a hardcoded `snapshots.reth.rs` URL that bypassed the `DownloadDefaults` override mechanism.

.changelog/fix-simulate-revert-code.md

@@ -0,0 +1,5 @@
+---
+reth-rpc-eth-types: patch
+---
+
+Updated `eth_simulateV1` revert error code from `-32000` to `3` to be consistent with `eth_call`, per [execution-apis#748](https://github.com/ethereum/execution-apis/pull/748).

.changelog/icy-lions-slide.md

@@ -0,0 +1,7 @@
+---
+reth: patch
+reth-cli-commands: patch
+reth-node-core: patch
+---
+
+Removed experimental ress protocol support for stateless Ethereum nodes.

.changelog/icy-tigers-buzz.md

@@ -0,0 +1,6 @@
+---
+reth-trie-db: minor
+reth-engine-tree: minor
+---
+
+Added `PendingChangeset` and `PendingChangesetGuard` to `ChangesetCache` so concurrent readers wait for an in-progress computation instead of falling back to the expensive DB-based path. The guard automatically cancels the pending entry on drop (e.g. task panic), ensuring waiters always make progress.

.changelog/keen-geese-bake.md

@@ -0,0 +1,5 @@
+---
+reth-engine-tree: patch
+---
+
+Added sub-phase timing histograms to the sparse trie event loop, tracking channel wait, proof coalescing, multiproof reveal, and trie update durations separately.

.changelog/kind-eagles-hum.md

@@ -0,0 +1,6 @@
+---
+reth-primitives-traits: major
+reth-downloaders: patch
+---
+
+Removed the local `size` module from `reth-primitives-traits` and replaced it with `alloy_consensus::InMemorySize`. Simplified `SignedTransaction` to a blanket impl covering all types satisfying the required bounds, removing `is_system_tx`, `auto_impl` attributes, and explicit impls for `EthereumTxEnvelope` and OP types. Updated import paths in `reth-downloaders` accordingly.

.changelog/kind-frogs-meow.md

@@ -0,0 +1,13 @@
+---
+reth-primitives-traits: minor
+reth-engine-local: patch
+reth-evm: patch
+reth-node-builder: patch
+reth-payload-primitives: patch
+reth-rpc-convert: patch
+reth-rpc-eth-api: patch
+reth-db-api: patch
+reth-db: patch
+---
+
+Removed the unused `Extended` type and `op` feature (including `op-alloy-consensus` dependency) from `reth-primitives-traits`. Updated all dependent crates to remove the now-unnecessary `reth-primitives-traits/op` feature flag propagation.

.changelog/lazy-lakes-shout.md

@@ -0,0 +1,5 @@
+---
+reth-node-builder: patch
+---
+
+Removed biased select in engine service loop to allow fair scheduling of shutdown requests alongside event processing.

.changelog/lively-clams-drink.md

@@ -0,0 +1,5 @@
+---
+reth-transaction-pool: patch
+---
+
+Fixed swapped arguments in `blob_tx_priority` function calls, correcting the parameter order to match the function signature.

.changelog/lively-clouds-bake.md

@@ -0,0 +1,4 @@
+---
+---
+
+Improved documentation overview page with better structure and clarity.

.changelog/lively-deer-shout.md

@@ -0,0 +1,5 @@
+---
+reth: patch
+---
+
+Re-enabled changelog workflow to run automatically on pull requests.

.changelog/lively-foxes-play.md

@@ -0,0 +1,5 @@
+---
+reth-node-events: patch
+---
+
+Updated consensus engine log message to be more accurate about received updates.

.changelog/merry-bees-break.md

@@ -0,0 +1,8 @@
+---
+reth-engine-primitives: minor
+reth-engine-tree: minor
+reth-node-core: minor
+reth-trie-parallel: minor
+---
+
+Added `--engine.proof-jitter` CLI option behind the `trie-debug` feature flag. When set, each proof worker sleeps for a random duration up to the specified value before starting proof computation, useful for stress-testing timing-sensitive proof logic.

.changelog/merry-koalas-nod.md

@@ -0,0 +1,6 @@
+---
+reth-rpc-eth-api: minor
+reth-rpc-server-types: minor
+---
+
+Added `eth_getStorageValues` RPC method for batch storage slot retrieval across multiple addresses.

.changelog/mild-foxes-bark.md

@@ -0,0 +1,6 @@
+---
+reth-chainspec: minor
+reth-network-peers: minor
+---
+
+Removed OP stack bootnodes from default chain configurations and network peers module.

.changelog/mild-foxes-cook.md

@@ -0,0 +1,5 @@
+---
+reth-rpc-convert: patch
+---
+
+Updated `alloy-evm` dependency to git revision `9bc2dba` and adapted `TxEnvConverter` impl to the updated `TryIntoTxEnv` trait signature that now includes a `Spec` generic parameter.

.changelog/neat-ducks-whisper.md

@@ -0,0 +1,6 @@
+---
+reth-trie: patch
+reth-trie-sparse: patch
+---
+
+Refactored test harness for sparse trie tests by extracting `TrieTestHarness` into a shared `reth-trie` test utility, replacing duplicated inline harness code across multiple test modules. Updated `proof_v2` return type to include an optional root hash, and converted `original_root` and `storage` from public fields to accessor methods.

.changelog/nice-bears-cook.md

@@ -0,0 +1,5 @@
+---
+reth-trie: patch
+---
+
+Removed the local `increment_and_strip_trailing_zeros` function and `PATH_ALL_ZEROS` static in `proof_v2`, replacing them with the equivalent `Nibbles::next_without_prefix` and `Nibbles::is_zeroes` builtins. Also replaced manual `.get()` calls on `state_mask`/`hash_mask` with direct field access and switched to `Nibbles::unpack_array` over the unsafe `unpack_unchecked`.

.changelog/nice-trees-drink.md

@@ -0,0 +1,9 @@
+---
+reth-network-api: minor
+reth-network-types: minor
+reth-network: minor
+reth-node-core: minor
+reth: minor
+---
+
+Added optional ENR fork ID enforcement to filter out peers from incompatible networks during peer discovery, controlled by the `--enforce-enr-fork-id` CLI flag.

.changelog/nice-waves-bow.md

@@ -0,0 +1,5 @@
+---
+reth-primitives: patch
+---
+
+Moved feature-referenced dependencies from dev-dependencies to optional dependencies to ensure they are available when their corresponding features are enabled.

.changelog/nice-whales-buzz.md

@@ -0,0 +1,5 @@
+---
+reth-engine-tree: patch
+---
+
+Added idle-time pre-computation of account trie upper hashes in the sparse trie payload processor when no pending proof results are available.

.changelog/odd-birds-meow.md

@@ -0,0 +1,6 @@
+---
+reth-engine-local: patch
+reth-node-builder: patch
+---
+
+Removed the `op` feature flag and `OpPayloadAttributes` `PayloadAttributesBuilder` implementation from `reth-engine-local`, along with the `op-alloy-rpc-types-engine` dependency. Updated `reth-node-builder` to no longer enable the removed `op` feature on `reth-engine-local`.

.changelog/odd-donkeys-chirp.md

@@ -0,0 +1,5 @@
+---
+reth-engine-tree: patch
+---
+
+Fixed `compare_trie_updates` to return `bool` indicating whether differences were found, and updated the caller to properly use the return value instead of treating all successful comparisons as having no differences.

.changelog/odd-frogs-break.md

@@ -0,0 +1,7 @@
+---
+reth-cli-commands: minor
+reth-node-core: minor
+reth: patch
+---
+
+Made v2 storage the default for all new databases, deprecating the `--storage.v2` flag to a hidden no-op kept for backwards compatibility. Updated CLI reference docs to remove the now-hidden flag from all command help pages.

.changelog/odd-snails-play.md

@@ -0,0 +1,5 @@
+---
+reth-node-core: minor
+---
+
+Added `with_dev_block_time` helper method to `NodeConfig` for configuring dev miner block production interval.

.changelog/old-dogs-shake.md

@@ -0,0 +1,5 @@
+---
+reth-transaction-pool: minor
+---
+
+Added `IntoIter: Send` bounds to `validate_transactions` and `validate_transactions_with_origin` in the `TransactionValidator` trait, avoiding unnecessary `Vec` collects. Simplified default `validate_transactions_with_origin` to delegate to `validate_transactions`.

.changelog/proud-crabs-bark.md

@@ -0,0 +1,5 @@
+---
+reth-db-api: patch
+---
+
+Changed `StoredNibblesSubKey` encoding to use a stack-allocated `[u8; 65]` array instead of a heap-allocated `Vec<u8>`, avoiding unnecessary heap allocation.

.changelog/proud-slugs-fly.md

@@ -0,0 +1,7 @@
+---
+reth-engine-tree: patch
+reth-trie-sparse: patch
+reth-tasks: patch
+---
+
+Offloaded deallocation of expensive proof node buffers to a persistent background thread (`Runtime::spawn_drop`) to avoid blocking state root computation or lock-holding code.

.changelog/proud-wolves-spin.md

@@ -0,0 +1,5 @@
+---
+reth-storage-api: patch
+---
+
+Added `Arc` to `auto_impl` derive for storage-api traits to support automatic `Arc` wrapper implementations.

.changelog/quick-mice-jump.md

@@ -0,0 +1,8 @@
+---
+reth: patch
+reth-engine-tree: patch
+reth-node-builder: patch
+reth-trie-sparse: minor
+---
+
+Added `trie-debug` feature for recording sparse trie mutations to aid in debugging state root mismatches.

.changelog/quiet-foxes-dance.md

@@ -0,0 +1,6 @@
+---
+reth-static-file-types: patch
+reth-provider: patch
+---
+
+Move changeset offsets from segment header to external `.csoff` sidecar file for incremental writes and crash recovery.

.changelog/quiet-frogs-whisper.md

@@ -0,0 +1,5 @@
+---
+reth-provider: patch
+---
+
+Removed unused staging types from ProviderFactoryBuilder.

.changelog/rare-tigers-nod.md

@@ -0,0 +1,6 @@
+---
+reth-trie-sparse: patch
+reth-engine-tree: patch
+---
+
+Removed the `skip_proof_node_filtering` flag, `revealed_account_paths`/`revealed_paths` tracking, and the `filter_revealed_v2_proof_nodes` function from the sparse trie implementation. Also removed the corresponding skipped-nodes metrics, simplifying the proof node reveal path to always pass nodes directly to the sparse trie without pre-filtering.

.changelog/rare-whales-pack.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: minor
+---
+
+Added a comprehensive generic `SparseTrie` test suite covering `set_root`, `reveal_nodes`, `update_leaves`, `root`, `take_updates`, `commit_updates`, `prune`, `wipe`/`clear`, `get_leaf_value`, `find_leaf`, `size_hint`, and integration lifecycle scenarios. Tests are stamped out for all concrete `SparseTrie` implementations via a macro.

.changelog/rich-lakes-bark.md

@@ -0,0 +1,5 @@
+---
+reth-provider: patch
+---
+
+Fixed sender pruning during block reorg to skip when sender_recovery is fully pruned, preventing a fatal crash when no sender data exists in static files.

.changelog/rich-pigs-shake.md

@@ -0,0 +1,6 @@
+---
+reth-payload-primitives: patch
+reth-engine-local: patch
+---
+
+Removed the `op` feature and `op-alloy-rpc-types-engine` dependency from `reth-payload-primitives`, along with the `ExecutionPayload` impl for `OpExecutionData`. Updated `reth-engine-local` to drop the corresponding feature flag dependency.

.changelog/safe-waves-cook.md

@@ -0,0 +1,5 @@
+---
+reth-engine-tree: patch
+---
+
+Downgraded per-transaction prewarm span from `debug_span!` to `trace_span!` to reduce noise in debug-level logging.

.changelog/safe-waves-read.md

@@ -0,0 +1,7 @@
+---
+reth-network-types: minor
+reth-network: minor
+reth-node-core: patch
+---
+
+Added `PersistedPeerInfo` struct to persist richer peer metadata (kind, fork ID, reputation) to disk. Updated `PeersConfig::with_basic_nodes_from_file` to support both the new `PersistedPeerInfo` format and the legacy `Vec<NodeRecord>` format with automatic conversion, and updated `write_peers_to_file` to exclude backed-off and banned peers.

.changelog/shy-tigers-dry.md

@@ -1,6 +1,5 @@
 ---
 reth: patch
-op-reth: patch
 ---
 
 Added automated changelog generation infrastructure using wevm/changelogs-rs with Claude Code integration. Configured per-crate changelog format with fixed version groups for reth binaries and exclusions for internal test utilities.

.changelog/swift-owls-fly.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: minor
+---
+
+Removed `SerialSparseTrie` from the workspace, consolidating on `ParallelSparseTrie` as the single sparse trie implementation in `reth-trie-sparse`.

.changelog/tall-fish-climb.md

@@ -0,0 +1,5 @@
+---
+reth-cli-commands: minor
+---
+
+Added `reth_version` field to `SnapshotManifest` to record the Reth version that produced a snapshot. The field is optional and populated automatically during manifest generation.

.changelog/tall-stars-shout.md

@@ -0,0 +1,5 @@
+---
+reth-network: minor
+---
+
+Added `fork_id` as a tiebreaker in peer selection when reputations are equal, preferring peers with a discovered `fork_id` as it indicates fork compatibility. Added a test to verify the tiebreaker behavior.

.changelog/tidy-ducks-bake.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: patch
+---
+
+Fixed a panic in `ParallelSparseTrie::reveal_nodes` when a boundary node's upper parent is absent or non-branch (e.g. when an upper extension crosses the boundary). The code now skips gracefully instead of unwrapping. Added a regression test covering this case.

.changelog/tidy-stars-cry.md

@@ -0,0 +1,5 @@
+---
+reth-trie-sparse: patch
+---
+
+Fixed a bug where trie nodes could appear in both `updated_nodes` and `removed_nodes` simultaneously by removing entries from `removed_nodes` when a node is inserted as updated.

.changelog/vain-bats-bark.md

@@ -0,0 +1,8 @@
+---
+reth-trie-common: minor
+reth-trie: minor
+reth-trie-parallel: minor
+reth-engine-tree: patch
+---
+
+Moved `ProofV2Target`, `MultiProofTargetsV2`, and `ChunkedMultiProofTargetsV2` from `reth-trie-parallel::targets_v2` into a new `reth-trie-common::target_v2` module, making these types available at a lower level without pulling in the full parallel trie crate. Added a `multiproof_v2` method to `Proof` in `reth-trie` that generates a state multiproof using the V2 proof calculator with synchronous account value encoding.

.changelog/vain-birds-laugh.md

@@ -0,0 +1,14 @@
+---
+reth-engine-tree: patch
+reth-evm-ethereum: patch
+reth-evm: patch
+reth-primitives-traits: patch
+reth-revm: patch
+reth-rpc-eth-api: patch
+reth-rpc-eth-types: patch
+reth-provider: patch
+example-custom-evm: patch
+example-precompile-cache: patch
+---
+
+Bumped revm to v35.0.0, revm-inspectors to 0.35.0, and alloy-evm to 0.29.0. Updated call sites throughout the codebase to align with the new APIs, including `ExecutionResult` field changes (`gas_used` → `gas.used()`/`gas.final_refunded()`), removal of `.without_state_clear()`, updated `EthPrecompiles::new(spec)` constructor, and updated `block_hashes.lowest()` access.

.changelog/vain-goats-cook.md

@@ -0,0 +1,9 @@
+---
+reth-trie-sparse: minor
+reth-engine-primitives: minor
+reth-engine-tree: minor
+reth-node-core: minor
+reth-trie-common: patch
+---
+
+Added an arena-based sparse trie implementation (`ArenaParallelSparseTrie`) using `slotmap` arena allocation for node storage, enabling parallel subtrie mutation without per-node hashing overhead. Added `ConfigurableSparseTrie` enum to switch between the arena and hash-map implementations, and a `--engine.enable-arena-sparse-trie` CLI flag to opt in at runtime.

.changelog/vast-waves-fold.md

@@ -0,0 +1,4 @@
+---
+---
+
+Expanded CLI integration tests with subcommand help coverage, config TOML validation, genesis JSON validation, and send transaction round-trip test for dev mode.

.changelog/warm-donkeys-dry.md

@@ -0,0 +1,6 @@
+---
+reth-engine-local: minor
+reth-node-builder: minor
+---
+
+Added trigger-based `MiningMode` variant that allows blocks to be built on-demand via custom streams, and exposed `with_mining_mode` method on `DebugNodeLauncherFuture` to override default mining configuration.

.changelog/warm-foxes-glow.md

@@ -0,0 +1,5 @@
+---
+reth-network: minor
+---
+
+Added direction labels to `closed_sessions` and `pending_session_failures` metrics. Operators can now distinguish session closures and failures by direction (`active`, `incoming_pending`, `outgoing_pending` for closed sessions; `inbound`, `outbound` for pending session failures).

.changelog/warm-foxes-jump.md

@@ -0,0 +1,4 @@
+---
+---
+
+Moved Kurtosis CI failure notifications to the hive Slack channel.

.changelog/warm-geese-build.md

@@ -0,0 +1,7 @@
+---
+reth-rpc-api: minor
+reth-rpc-builder: patch
+reth-rpc: minor
+---
+
+Added `subscribeFinalizedChainNotifications` RPC endpoint that buffers committed chain notifications and emits them once a new finalized block is received.

.changelog/warm-pandas-cook.md

@@ -0,0 +1,5 @@
+---
+reth-transaction-pool: patch
+---
+
+Fixed a bug where transactions from the same sender were added to the pending subpool out of nonce order. Ensured `process_updates` runs before `add_new_transaction` so that lower-nonce promotions are enqueued before the newly inserted higher-nonce transaction, preserving correct ordering for live `BestTransactions` iterators.

.changelog/warm-slugs-crawl.md

@@ -0,0 +1,5 @@
+---
+reth-transaction-pool: minor
+---
+
+Added `TransactionValidationTaskExecutor::spawn` as a dedicated constructor that encapsulates spawning validation tasks on a runtime, and refactored `EthTransactionValidatorBuilder::build_with_tasks` to use it.

.changelog/warm-trees-grow.md

@@ -0,0 +1,7 @@
+---
+reth-engine-primitives: patch
+reth-engine-tree: patch
+reth-node-core: patch
+---
+
+Removed `--engine.enable-arena-sparse-trie` CLI flag and made the arena-based sparse trie the default implementation. The hash-map-based `ParallelSparseTrie` variant is no longer selectable.

.changelog/young-horses-play.md

@@ -0,0 +1,7 @@
+---
+reth-trie: major
+reth-trie-db: major
+reth-provider: minor
+---
+
+Added `MaskedTrieCursorFactory` and `MaskedTrieCursor` to handle prefix-set-based hash invalidation at the cursor layer, replacing the `DatabaseTrieWitness` trait abstraction. Removed `with_prefix_sets_mut` from `TrieWitness` and deleted `DatabaseTrieWitness` — callers should now wrap their cursor factory with `MaskedTrieCursorFactory` to apply prefix sets during witness/proof computation.

.changelog/zesty-bees-drink.md

@@ -0,0 +1,8 @@
+---
+reth-engine-primitives: minor
+reth-engine-tree: major
+reth-node-core: minor
+reth-cli-commands: minor
+---
+
+Added persistence backpressure to the engine tree: when the canonical-minus-persisted block gap exceeds a configurable threshold (`--engine.persistence-backpressure-threshold`, default 16), the engine loop stalls on the persistence receiver instead of processing new incoming messages. Added CLI argument, cross-field validation, metrics (`backpressure_active`, `backpressure_stall_duration`), and tests.

.changelog/zesty-birds-smile.md

@@ -0,0 +1,6 @@
+---
+reth-trie: minor
+reth-trie-parallel: minor
+---
+
+Added `root_node` and `storage_root_node` methods to proof calculators for efficient root-only calculations. These methods directly return the root node without requiring dummy targets, replacing the previous workaround of passing fake targets to proof generation.

.changelog/zesty-clouds-wave.md

@@ -0,0 +1,5 @@
+---
+reth-trie: patch
+---
+
+Fixed a potential panic in `ProofCalculator` by clearing internal computation state (`branch_stack`, `child_stack`, `branch_path`, etc.) after errors, preventing stale state from causing `usize` underflow panics when the calculator is reused. Added a test verifying correct behavior after simulated mid-computation errors.

.config/nextest.toml

@@ -1,6 +1,6 @@
 [profile.default]
 retries = { backoff = "exponential", count = 2, delay = "2s", jitter = true }
-slow-timeout = { period = "30s", terminate-after = 4 }
+slow-timeout = { period = "30s", terminate-after = 2 }
 
 [[profile.default.overrides]]
 filter = "test(general_state_tests)"

.config/zepter.yaml

@@ -12,7 +12,7 @@ workflows:
         # Check that `A` activates the features of `B`.
         "propagate-feature",
         # These are the features to check:
-        "--features=std,op,dev,asm-keccak,jemalloc,jemalloc-prof,tracy-allocator,tracy,serde-bincode-compat,serde,test-utils,arbitrary,bench,alloy-compat,min-error-logs,min-warn-logs,min-info-logs,min-debug-logs,min-trace-logs,otlp,otlp-logs,js-tracer,portable,keccak-cache-global",
+        "--features=std,op,dev,asm-keccak,jemalloc,jemalloc-prof,tracy-allocator,tracy,serde-bincode-compat,serde,test-utils,arbitrary,bench,alloy-compat,min-error-logs,min-warn-logs,min-info-logs,min-debug-logs,min-trace-logs,otlp,otlp-logs,js-tracer,portable,keccak-cache-global,trie-debug",
         # Do not try to add a new section to `[features]` of `A` only because `B` exposes that feature. There are edge-cases where this is still needed, but we can add them manually.
         "--left-side-feature-missing=ignore",
         # Ignore the case that `A` it outside of the workspace. Otherwise it will report errors in external dependencies that we have no influence on.

.dockerignore

@@ -20,6 +20,11 @@
 # include dist directory, where the reth binary is located after compilation
 !/dist
 
+# include PGO build helper used by Dockerfile.depot
+!/.github
+!/.github/scripts
+!/.github/scripts/build_pgo_bolt.sh
+
 # include licenses
 !LICENSE-*
 

.github/CODEOWNERS

@@ -1,7 +1,7 @@
 *                           @gakonst
 crates/chain-state/         @fgimenez @mattsse
 crates/chainspec/           @Rjected @joshieDo @mattsse
-crates/cli/                 @mattsse
+crates/cli/                 @mattsse @Rjected
 crates/config/              @shekhirin @mattsse @Rjected 
 crates/consensus/           @mattsse @Rjected
 crates/e2e-test-utils/      @mattsse @Rjected @klkvr @fgimenez
@@ -19,7 +19,6 @@ crates/metrics/             @mattsse @Rjected
 crates/net/                 @mattsse @Rjected
 crates/net/downloaders/     @Rjected
 crates/node/                @mattsse @Rjected @klkvr
-crates/optimism/            @mattsse @Rjected
 crates/payload/             @mattsse @Rjected
 crates/primitives-traits/   @Rjected @mattsse @klkvr
 crates/primitives/          @Rjected @mattsse @klkvr
@@ -39,7 +38,7 @@ crates/storage/libmdbx-rs/  @shekhirin
 crates/storage/nippy-jar/   @joshieDo @shekhirin
 crates/storage/provider/    @joshieDo @shekhirin @yongkangc
 crates/storage/storage-api/ @joshieDo
-crates/tasks/               @mattsse
+crates/tasks/               @mattsse @DaniPopes
 crates/tokio-util/          @mattsse 
 crates/tracing/             @mattsse @shekhirin
 crates/tracing-otlp/        @mattsse @Rjected 

.github/ISSUE_TEMPLATE/bug.yml

@@ -43,7 +43,6 @@ body:
 
         - `~/.cache/reth/logs` on Linux
         - `~/Library/Caches/reth/logs` on macOS
-        - `%localAppData%/reth/logs` on Windows
       render: text
     validations:
       required: false
@@ -58,8 +57,6 @@ body:
         - Linux (ARM)
         - Mac (Intel)
         - Mac (Apple Silicon)
-        - Windows (x86)
-        - Windows (ARM)
   - type: dropdown
     id: container_type
     attributes:

.github/actionlint.yaml

@@ -5,3 +5,4 @@ self-hosted-runner:
     - depot-ubuntu-latest-4
     - depot-ubuntu-latest-8
     - depot-ubuntu-latest-16
+    - available

.github/assets/kurtosis_op_network_params.yaml

@@ -1,36 +0,0 @@
-ethereum_package:
-  participants:
-    - el_type: reth
-      el_extra_params:
-        - "--rpc.eth-proof-window=100"
-      cl_type: teku
-  network_params:
-    preset: minimal
-    genesis_delay: 5
-    additional_preloaded_contracts: '
-      {
-        "0x4e59b44847b379578588920cA78FbF26c0B4956C": {
-          "balance": "0ETH",
-          "code": "0x7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffe03601600081602082378035828234f58015156039578182fd5b8082525050506014600cf3",
-          "storage": {},
-          "nonce": "1"
-        }
-      }'
-optimism_package:
-  chains:
-    chain0:
-      participants:
-        node0:
-          el:
-            type: op-geth
-          cl:
-            type: op-node
-        node1:
-          el:
-            type: op-reth
-            image: "ghcr.io/paradigmxyz/op-reth:kurtosis-ci"
-          cl:
-            type: op-node
-      network_params:
-        holocene_time_offset: 0
-        isthmus_time_offset: 0

.github/dependabot.yml

@@ -4,3 +4,17 @@ updates:
   directory: "/"
   schedule:
     interval: "weekly"
+- package-ecosystem: "cargo"
+  directory: "/"
+  schedule:
+    interval: "weekly"
+  labels:
+    - "A-dependencies"
+  commit-message:
+    prefix: "chore(deps)"
+  open-pull-requests-limit: 1
+  groups:
+    cargo-weekly:
+      applies-to: "version-updates"
+      patterns: ["*"]
+      update-types: ["minor", "patch"]

.github/scripts/bench-job-summary.js

@@ -0,0 +1,106 @@
+// Generates a rich GitHub Actions job summary for reth-bench results.
+//
+// Reads from environment:
+//   BENCH_WORK_DIR       – Directory containing summary.json
+//   BENCH_PR             – PR number (may be empty)
+//   BENCH_ACTOR          – GitHub user who triggered the bench
+//   BENCH_CORES          – CPU core limit (0 = all)
+//   BENCH_WARMUP_BLOCKS  – Number of warmup blocks
+//   BENCH_SAMPLY         – 'true' if samply profiling was enabled
+//   BENCH_ABBA           – 'true' if ABBA interleaved order was used
+//
+// Usage from actions/github-script:
+//   const jobSummary = require('./.github/scripts/bench-job-summary.js');
+//   await jobSummary({ core, context, chartSha, grafanaUrl, runId });
+
+const fs = require('fs');
+const { verdict, loadSamplyUrls, blocksLabel, metricRows, waitTimeRows } = require('./bench-utils');
+
+module.exports = async function ({ core, context, chartSha, grafanaUrl, runId }) {
+  let summary;
+  try {
+    summary = JSON.parse(fs.readFileSync(process.env.BENCH_WORK_DIR + '/summary.json', 'utf8'));
+  } catch (e) {
+    await core.summary.addRaw('⚠️ Benchmark completed but failed to load summary.').write();
+    return;
+  }
+
+  const repo = `${context.repo.owner}/${context.repo.repo}`;
+  const prNumber = process.env.BENCH_PR;
+  const actor = process.env.BENCH_ACTOR;
+  const commitUrl = `https://github.com/${repo}/commit`;
+
+  const { emoji, label } = verdict(summary.changes);
+  const baselineLink = `[\`${summary.baseline.name}\`](${commitUrl}/${summary.baseline.ref})`;
+  const featureLink = `[\`${summary.feature.name}\`](${commitUrl}/${summary.feature.ref})`;
+  const diffUrl = `https://github.com/${repo}/compare/${summary.baseline.ref}...${summary.feature.ref}`;
+
+  // Header & metadata
+  const metaParts = [];
+  if (prNumber) metaParts.push(`**[PR #${prNumber}](https://github.com/${repo}/pull/${prNumber})**`);
+  metaParts.push(`triggered by @${actor}`);
+
+  let md = `# ${emoji} ${label}\n\n`;
+  md += metaParts.join(' · ') + '\n\n';
+  md += `**Baseline:** ${baselineLink}\n`;
+  md += `**Feature:** ${featureLink} ([diff](${diffUrl}))\n`;
+  md += blocksLabel(summary).map(p => `**${p.key}:** ${p.value}`).join(' · ') + '\n\n';
+
+  // Main comparison table
+  const rows = metricRows(summary);
+  md += `| Metric | Baseline | Feature | Change |\n`;
+  md += `|--------|----------|---------|--------|\n`;
+  for (const r of rows) {
+    md += `| ${r.label} | ${r.baseline} | ${r.feature} | ${r.change} |\n`;
+  }
+  md += '\n';
+
+  // Wait time breakdown
+  const wtRows = waitTimeRows(summary);
+  if (wtRows.length > 0) {
+    md += `### Wait Time Breakdown\n\n`;
+    md += `| Metric | Baseline | Feature |\n`;
+    md += `|--------|----------|--------|\n`;
+    for (const r of wtRows) {
+      md += `| ${r.title} | ${r.baseline} | ${r.feature} |\n`;
+    }
+    md += '\n';
+  }
+
+  // Charts
+  if (chartSha) {
+    const prNum = prNumber || '0';
+    const baseUrl = `https://raw.githubusercontent.com/decofe/reth-bench-charts/${chartSha}/pr/${prNum}/${runId}`;
+    const charts = [
+      { file: 'latency_throughput.png', label: 'Latency, Throughput & Diff' },
+      { file: 'wait_breakdown.png', label: 'Wait Time Breakdown' },
+      { file: 'gas_vs_latency.png', label: 'Gas vs Latency' },
+    ];
+    md += `### Charts\n\n`;
+    for (const chart of charts) {
+      md += `<details><summary>${chart.label}</summary>\n\n`;
+      md += `![${chart.label}](${baseUrl}/${chart.file})\n\n`;
+      md += `</details>\n\n`;
+    }
+  }
+
+  // Samply profiles
+  const samplyUrls = loadSamplyUrls(process.env.BENCH_WORK_DIR);
+  const samplyLinks = Object.entries(samplyUrls).map(([run, url]) => `- **${run}**: [Firefox Profiler](${url})`);
+  if (samplyLinks.length > 0) {
+    md += `### Samply Profiles\n\n${samplyLinks.join('\n')}\n\n`;
+  }
+
+  // Grafana
+  if (grafanaUrl) {
+    md += `### Grafana Dashboard\n\n[View real-time metrics](${grafanaUrl})\n\n`;
+  }
+
+  // Node errors
+  try {
+    const errors = fs.readFileSync(process.env.BENCH_WORK_DIR + '/errors.md', 'utf8');
+    if (errors.trim()) md += '\n' + errors + '\n';
+  } catch {}
+
+  await core.summary.addRaw(md).write();
+};

.github/scripts/bench-metrics-proxy.py

@@ -0,0 +1,276 @@
+#!/usr/bin/env python3
+"""
+Prometheus metrics proxy that fetches from a local reth node and
+re-exposes with additional benchmark labels.
+
+Reads labels from a JSON file (updated by local-reth-bench.sh between runs)
+and injects them into every Prometheus metric line.
+
+Returns empty 200 when reth is not running (clean Grafana gaps).
+"""
+import argparse
+import ipaddress
+import json
+import subprocess
+import sys
+import time
+from http.server import HTTPServer, BaseHTTPRequestHandler
+from urllib.request import urlopen
+from urllib.error import URLError
+
+
+def read_labels(path):
+    try:
+        with open(path) as f:
+            return json.load(f)
+    except (FileNotFoundError, json.JSONDecodeError):
+        return {}
+
+
+def inject_labels(metrics_bytes, label_str, label_names):
+    """Inject labels into Prometheus text format.
+
+    Operates on bytes and uses simple string ops instead of regex
+    for speed on large payloads (reth exposes thousands of metrics).
+
+    Skips injecting into lines that already contain any of the label names
+    to avoid duplicate labels (which Prometheus rejects).
+    """
+    if not label_str:
+        return metrics_bytes
+
+    label_bytes = label_str.encode("utf-8")
+    # Pre-encode label names for fast duplicate detection
+    label_name_bytes = [n.encode("utf-8") for n in label_names]
+    out = []
+    for line in metrics_bytes.split(b"\n"):
+        # Skip comments and blank lines
+        if line.startswith(b"#") or not line:
+            out.append(line)
+            continue
+
+        brace = line.find(b"{")
+        space = line.find(b" ")
+
+        if space == -1:
+            # Malformed, pass through
+            out.append(line)
+        elif brace != -1 and brace < space:
+            # Has labels: metric{existing="val"} 123
+            close = line.find(b"}", brace)
+            if close == -1:
+                out.append(line)
+                continue
+
+            # Filter out labels that already exist in this line
+            existing = line[brace + 1:close]
+            inject = label_bytes
+            if existing:
+                for name in label_name_bytes:
+                    if name + b"=" in existing:
+                        # Rebuild inject string excluding this label
+                        inject = _remove_label(inject, name)
+                if not inject:
+                    out.append(line)
+                    continue
+
+            if close == brace + 1:
+                # Empty braces: metric{} 123
+                out.append(line[:close] + inject + line[close:])
+            else:
+                out.append(line[:close] + b"," + inject + line[close:])
+        else:
+            # No labels: metric 123
+            out.append(line[:space] + b"{" + label_bytes + b"}" + line[space:])
+
+    return b"\n".join(out)
+
+
+def _remove_label(label_bytes, name):
+    """Remove a single label (name=\"...\") from a comma-separated label string."""
+    parts = []
+    for part in label_bytes.split(b","):
+        if not part.startswith(name + b"="):
+            parts.append(part)
+    return b",".join(parts)
+
+
+def build_label_str(labels):
+    """Pre-format the label injection string: key1="val1",key2="val2" """
+    if not labels:
+        return ""
+    return ",".join(f'{k}="{v}"' for k, v in sorted(labels.items()))
+
+
+def build_elapsed_gauge(labels):
+    """Build a bench_elapsed_seconds gauge from run_start_epoch in labels."""
+    start = labels.get("run_start_epoch")
+    if not start:
+        return b""
+    try:
+        elapsed = time.time() - float(start)
+    except (ValueError, TypeError):
+        return b""
+    # Build labels excluding internal keys
+    display = {k: v for k, v in labels.items()
+               if k not in ("run_start_epoch", "reference_epoch")}
+    lstr = build_label_str(display)
+    return (
+        f"# HELP bench_elapsed_seconds Seconds since benchmark run started\n"
+        f"# TYPE bench_elapsed_seconds gauge\n"
+        f"bench_elapsed_seconds{{{lstr}}} {elapsed:.1f}\n"
+    ).encode("utf-8")
+
+
+def compute_timestamp_ms(labels):
+    """Compute a synthetic timestamp so all runs share a common time origin.
+
+    Returns the timestamp in milliseconds, or None if not enough info.
+    Uses: reference_epoch + (now - run_start_epoch) → all runs overlay at
+    the same Grafana time range.
+    """
+    ref = labels.get("reference_epoch")
+    start = labels.get("run_start_epoch")
+    if not ref or not start:
+        return None
+    try:
+        elapsed = time.time() - float(start)
+        return int((float(ref) + elapsed) * 1000)
+    except (ValueError, TypeError):
+        return None
+
+
+def inject_timestamps(metrics_bytes, timestamp_ms):
+    """Append a Prometheus timestamp (ms) to every data line.
+
+    Prometheus text format: metric{labels} value [timestamp_ms]
+    Adding timestamps causes Prometheus to store all runs' samples
+    at the same relative time, enabling natural overlay in Grafana.
+    """
+    if timestamp_ms is None:
+        return metrics_bytes
+
+    ts = str(timestamp_ms).encode("utf-8")
+    out = []
+    for line in metrics_bytes.split(b"\n"):
+        if line.startswith(b"#") or not line:
+            out.append(line)
+        else:
+            out.append(line + b" " + ts)
+    return b"\n".join(out)
+
+
+class MetricsHandler(BaseHTTPRequestHandler):
+    # Use HTTP/1.1 so Content-Length is respected and Prometheus
+    # doesn't have to rely on connection close to detect end of body.
+    protocol_version = "HTTP/1.1"
+
+    def do_GET(self):
+        src = self.client_address[0]
+        try:
+            resp = urlopen(self.server.upstream, timeout=2)
+            metrics = resp.read()
+        except (URLError, ConnectionError, OSError):
+            # reth not running — return empty 200
+            self._send(b"")
+            #print(f"  scrape from {src}: empty (reth not running)", flush=True)
+            return
+
+        all_labels = read_labels(self.server.labels_file)
+        # Internal keys — not injected as Prometheus labels
+        internal = ("run_start_epoch", "reference_epoch")
+        labels = {k: v for k, v in all_labels.items() if k not in internal}
+        label_str = build_label_str(labels)
+        label_names = sorted(labels.keys())
+
+        t0 = time.monotonic()
+        result = inject_labels(metrics, label_str, label_names)
+        result += build_elapsed_gauge(all_labels)
+        ts_ms = compute_timestamp_ms(all_labels)
+        result = inject_timestamps(result, ts_ms)
+        dt = time.monotonic() - t0
+
+        self._send(result)
+        print(f"  scrape from {src}: {len(metrics)} -> {len(result)} bytes, "
+              f"inject {dt*1000:.1f}ms", flush=True)
+
+    def _send(self, body):
+        self.send_response(200)
+        self.send_header("Content-Type", "text/plain; version=0.0.4")
+        self.send_header("Content-Length", str(len(body)))
+        self.send_header("Connection", "close")
+        self.end_headers()
+        if body:
+            self.wfile.write(body)
+
+    def log_message(self, format, *args):
+        pass  # suppress per-request logging
+
+
+def resolve_bind_address(subnet_cidr):
+    """Find the local IP address that belongs to the given subnet.
+
+    Uses ``ip -j addr show`` to enumerate interfaces and returns the first
+    address that falls within *subnet_cidr* (e.g. ``10.10.0.0/24``).
+    """
+    network = ipaddress.ip_network(subnet_cidr, strict=False)
+    try:
+        result = subprocess.run(
+            ["ip", "-j", "addr", "show"],
+            capture_output=True, text=True, check=True,
+        )
+        interfaces = json.loads(result.stdout)
+    except (subprocess.CalledProcessError, FileNotFoundError, json.JSONDecodeError) as exc:
+        print(f"Error: cannot enumerate interfaces: {exc}", file=sys.stderr)
+        sys.exit(1)
+
+    for iface in interfaces:
+        for addr_info in iface.get("addr_info", []):
+            try:
+                addr = ipaddress.ip_address(addr_info["local"])
+            except (KeyError, ValueError):
+                continue
+            if addr in network:
+                return str(addr)
+
+    print(f"Error: no interface address found in subnet {subnet_cidr}", file=sys.stderr)
+    sys.exit(1)
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Prometheus metrics proxy with label injection")
+    parser.add_argument("--labels", default="/tmp/bench-metrics-labels.json",
+                        help="Path to JSON file with labels to inject (default: /tmp/bench-metrics-labels.json)")
+    parser.add_argument("--upstream", default="http://127.0.0.1:9100/",
+                        help="Upstream reth metrics URL (default: http://127.0.0.1:9100/)")
+
+    bind_group = parser.add_mutually_exclusive_group()
+    bind_group.add_argument("--bind", default=None,
+                            help="Address to bind the proxy (default: 0.0.0.0)")
+    bind_group.add_argument("--subnet", default=None,
+                            help="Auto-detect bind address from a local interface in this subnet (e.g. 10.10.0.0/24)")
+
+    parser.add_argument("--port", type=int, default=9090,
+                        help="Port to bind the proxy (default: 9090)")
+    args = parser.parse_args()
+
+    if args.subnet:
+        bind_addr = resolve_bind_address(args.subnet)
+    elif args.bind:
+        bind_addr = args.bind
+    else:
+        bind_addr = "0.0.0.0"
+
+    server = HTTPServer((bind_addr, args.port), MetricsHandler)
+    server.upstream = args.upstream
+    server.labels_file = args.labels
+
+    print(f"bench-metrics-proxy listening on {bind_addr}:{args.port}")
+    print(f"  upstream: {args.upstream}")
+    print(f"  labels:   {args.labels}")
+    sys.stdout.flush()
+    server.serve_forever()
+
+
+if __name__ == "__main__":
+    main()

.github/scripts/bench-reth-build.sh

@@ -0,0 +1,144 @@
+#!/usr/bin/env bash
+#
+# Builds (or fetches from cache) reth binaries for benchmarking.
+#
+# Usage: bench-reth-build.sh <baseline|feature> <source-dir> <commit> [branch-sha]
+#
+#   baseline — build/fetch the baseline binary at <commit> (merge-base)
+#              source-dir must be checked out at <commit>
+#   feature  — build/fetch the candidate binary + reth-bench at <commit>
+#              source-dir must be checked out at <commit>
+#              optional branch-sha is the PR head commit for cache key
+#
+# Outputs:
+#   baseline: <source-dir>/target/profiling/reth (or reth-bb if BENCH_BIG_BLOCKS=true)
+#   feature:  <source-dir>/target/profiling/reth (or reth-bb), reth-bench installed to cargo bin
+#
+# Required: mc (MinIO client) with a configured alias
+# Optional env: BENCH_BIG_BLOCKS (true/false) — build reth-bb instead of reth
+set -euo pipefail
+
+MC="mc"
+MODE="$1"
+SOURCE_DIR="$2"
+COMMIT="$3"
+
+BIG_BLOCKS="${BENCH_BIG_BLOCKS:-false}"
+# The node binary to build: reth-bb for big blocks, reth otherwise
+if [ "$BIG_BLOCKS" = "true" ]; then
+  NODE_BIN="reth-bb"
+  NODE_PKG="-p reth-bb"
+else
+  NODE_BIN="reth"
+  NODE_PKG="--bin reth"
+fi
+
+# Tracy support: when BENCH_TRACY is "on" or "full", add Tracy cargo features
+# and frame pointers for accurate stack traces.
+EXTRA_FEATURES=""
+EXTRA_RUSTFLAGS=""
+if [ "${BENCH_TRACY:-off}" != "off" ]; then
+  EXTRA_FEATURES="tracy,tracy-client/ondemand"
+  EXTRA_RUSTFLAGS=" -C force-frame-pointers=yes"
+fi
+
+# Cache suffix: hash of features+rustflags so different build configs get separate cache entries
+if [ -n "$EXTRA_FEATURES" ] || [ -n "$EXTRA_RUSTFLAGS" ]; then
+  BUILD_SUFFIX="-$(echo "${EXTRA_FEATURES}${EXTRA_RUSTFLAGS}" | sha256sum | cut -c1-12)"
+else
+  BUILD_SUFFIX=""
+fi
+
+# Verify a cached reth binary was built from the expected commit.
+# `reth --version` outputs "Commit SHA: <full-sha>" on its own line.
+verify_binary() {
+  local binary="$1" expected_commit="$2"
+  local version binary_sha
+  version=$("$binary" --version 2>/dev/null) || return 1
+  binary_sha=$(echo "$version" | sed -n 's/^Commit SHA: *//p')
+  if [ -z "$binary_sha" ]; then
+    echo "Warning: could not extract commit SHA from version output"
+    return 1
+  fi
+  if [ "$binary_sha" = "$expected_commit" ]; then
+    return 0
+  fi
+  echo "Cache mismatch: binary built from ${binary_sha} but expected ${expected_commit}"
+  return 1
+}
+
+case "$MODE" in
+  baseline|main)
+    BUCKET="minio/reth-binaries/${COMMIT}${BUILD_SUFFIX}"
+    mkdir -p "${SOURCE_DIR}/target/profiling"
+
+    CACHE_VALID=false
+    if $MC stat "${BUCKET}/${NODE_BIN}" &>/dev/null; then
+      echo "Cache hit for baseline (${COMMIT}), downloading ${NODE_BIN}..."
+      $MC cp "${BUCKET}/${NODE_BIN}" "${SOURCE_DIR}/target/profiling/${NODE_BIN}"
+      chmod +x "${SOURCE_DIR}/target/profiling/${NODE_BIN}"
+      if verify_binary "${SOURCE_DIR}/target/profiling/${NODE_BIN}" "${COMMIT}"; then
+        CACHE_VALID=true
+      else
+        echo "Cached baseline binary is stale, rebuilding..."
+      fi
+    fi
+    if [ "$CACHE_VALID" = false ]; then
+      echo "Building baseline ${NODE_BIN} (${COMMIT}) from source..."
+      cd "${SOURCE_DIR}"
+      FEATURES_ARG=""
+      WORKSPACE_ARG=""
+      if [ -n "$EXTRA_FEATURES" ]; then
+        # --workspace is needed for cross-package feature syntax (tracy-client/ondemand)
+        FEATURES_ARG="--features ${EXTRA_FEATURES}"
+        WORKSPACE_ARG="--workspace"
+      fi
+      # shellcheck disable=SC2086
+      RUSTFLAGS="-C target-cpu=native${EXTRA_RUSTFLAGS}" \
+        cargo build --profile profiling $NODE_PKG $WORKSPACE_ARG $FEATURES_ARG
+      $MC cp "target/profiling/${NODE_BIN}" "${BUCKET}/${NODE_BIN}"
+    fi
+    ;;
+
+  feature|branch)
+    BRANCH_SHA="${4:-$COMMIT}"
+    BUCKET="minio/reth-binaries/${BRANCH_SHA}${BUILD_SUFFIX}"
+
+    CACHE_VALID=false
+    if $MC stat "${BUCKET}/${NODE_BIN}" &>/dev/null && $MC stat "${BUCKET}/reth-bench" &>/dev/null; then
+      echo "Cache hit for ${BRANCH_SHA}, downloading binaries..."
+      mkdir -p "${SOURCE_DIR}/target/profiling"
+      $MC cp "${BUCKET}/${NODE_BIN}" "${SOURCE_DIR}/target/profiling/${NODE_BIN}"
+      $MC cp "${BUCKET}/reth-bench" /home/ubuntu/.cargo/bin/reth-bench
+      chmod +x "${SOURCE_DIR}/target/profiling/${NODE_BIN}" /home/ubuntu/.cargo/bin/reth-bench
+      if verify_binary "${SOURCE_DIR}/target/profiling/${NODE_BIN}" "${COMMIT}"; then
+        CACHE_VALID=true
+      else
+        echo "Cached feature binary is stale, rebuilding..."
+      fi
+    fi
+    if [ "$CACHE_VALID" = false ]; then
+      echo "Building feature ${NODE_BIN} (${COMMIT}) from source..."
+      cd "${SOURCE_DIR}"
+      rustup show active-toolchain || rustup default stable
+      if [ -n "$EXTRA_FEATURES" ]; then
+        # Can't use `make profiling` when adding features; build explicitly
+        # --workspace is needed for cross-package feature syntax (tracy-client/ondemand)
+        RUSTFLAGS="-C target-cpu=native${EXTRA_RUSTFLAGS}" \
+          cargo build --profile profiling --workspace $NODE_PKG --features "${EXTRA_FEATURES}"
+      else
+        # shellcheck disable=SC2086
+        RUSTFLAGS="-C target-cpu=native${EXTRA_RUSTFLAGS}" \
+          cargo build --profile profiling $NODE_PKG
+      fi
+      make install-reth-bench
+      $MC cp "target/profiling/${NODE_BIN}" "${BUCKET}/${NODE_BIN}"
+      $MC cp "$(which reth-bench)" "${BUCKET}/reth-bench"
+    fi
+    ;;
+
+  *)
+    echo "Usage: $0 <baseline|feature> <source-dir> <commit> [branch-sha]"
+    exit 1
+    ;;
+esac

.github/scripts/bench-reth-charts.py

@@ -0,0 +1,260 @@
+#!/usr/bin/env python3
+"""Generate benchmark charts from reth-bench CSV output.
+
+Usage:
+    bench-engine-charts.py <combined_csv> --output-dir <dir> [--baseline <baseline_csv>]
+
+Generates three PNG charts:
+  1. newPayload latency + Ggas/s per block (+ latency diff when baseline present)
+  2. Wait breakdown (persistence, execution cache, sparse trie) per block
+  3. Scatter plot of gas used vs latency
+
+When --baseline is provided, charts overlay both datasets for comparison.
+"""
+
+import argparse
+import csv
+import sys
+from pathlib import Path
+
+import numpy as np
+
+try:
+    import matplotlib
+
+    matplotlib.use("Agg")
+    import matplotlib.pyplot as plt
+except ImportError:
+    print("matplotlib is required: pip install matplotlib", file=sys.stderr)
+    sys.exit(1)
+
+GIGAGAS = 1_000_000_000
+
+
+def parse_combined_csv(path: str) -> list[dict]:
+    rows = []
+    with open(path) as f:
+        reader = csv.DictReader(f)
+        for row in reader:
+            rows.append(
+                {
+                    "block_number": int(row["block_number"]),
+                    "gas_used": int(row["gas_used"]),
+                    "new_payload_latency_us": int(row["new_payload_latency"]),
+                    "persistence_wait_us": int(row["persistence_wait"])
+                    if row.get("persistence_wait")
+                    else None,
+                    "execution_cache_wait_us": int(row.get("execution_cache_wait", 0)),
+                    "sparse_trie_wait_us": int(row.get("sparse_trie_wait", 0)),
+                }
+            )
+    return rows
+
+
+def plot_latency_and_throughput(
+    feature: list[dict], baseline: list[dict] | None, out: Path,
+    baseline_name: str = "baseline", feature_name: str = "feature",
+):
+    num_plots = 3 if baseline else 2
+    fig, axes = plt.subplots(num_plots, 1, figsize=(12, 4 * num_plots), sharex=True)
+    ax1, ax2 = axes[0], axes[1]
+
+    feat_x = [r["block_number"] for r in feature]
+    feat_lat = [r["new_payload_latency_us"] / 1_000 for r in feature]
+    feat_ggas = []
+    for r in feature:
+        lat_s = r["new_payload_latency_us"] / 1_000_000
+        feat_ggas.append(r["gas_used"] / lat_s / GIGAGAS if lat_s > 0 else 0)
+
+    if baseline:
+        base_x = [r["block_number"] for r in baseline]
+        base_lat = [r["new_payload_latency_us"] / 1_000 for r in baseline]
+        base_ggas = []
+        for r in baseline:
+            lat_s = r["new_payload_latency_us"] / 1_000_000
+            base_ggas.append(r["gas_used"] / lat_s / GIGAGAS if lat_s > 0 else 0)
+        l, = ax1.plot(base_x, base_lat, linewidth=0.8, label=baseline_name, alpha=0.7)
+        ax1.axhline(np.median(base_lat), color=l.get_color(), linestyle="--", linewidth=1, alpha=0.7, label=f"{baseline_name} median")
+        l, = ax2.plot(base_x, base_ggas, linewidth=0.8, label=baseline_name, alpha=0.7)
+        ax2.axhline(np.median(base_ggas), color=l.get_color(), linestyle="--", linewidth=1, alpha=0.7, label=f"{baseline_name} median")
+
+    l, = ax1.plot(feat_x, feat_lat, linewidth=0.8, label=feature_name)
+    ax1.axhline(np.median(feat_lat), color=l.get_color(), linestyle="--", linewidth=1, label=f"{feature_name} median")
+    ax1.set_ylabel("Latency (ms)")
+    ax1.set_title("newPayload Latency per Block")
+    ax1.grid(True, alpha=0.3)
+    ax1.legend()
+
+    l, = ax2.plot(feat_x, feat_ggas, linewidth=0.8, label=feature_name)
+    ax2.axhline(np.median(feat_ggas), color=l.get_color(), linestyle="--", linewidth=1, label=f"{feature_name} median")
+    ax2.set_ylabel("Ggas/s")
+    ax2.set_title("Execution Throughput per Block")
+    ax2.grid(True, alpha=0.3)
+    ax2.legend()
+
+    if baseline:
+        ax3 = axes[2]
+        base_by_block = {r["block_number"]: r["new_payload_latency_us"] for r in baseline}
+        blocks, diffs = [], []
+        for r in feature:
+            bn = r["block_number"]
+            if bn in base_by_block and base_by_block[bn] > 0:
+                pct = (r["new_payload_latency_us"] - base_by_block[bn]) / base_by_block[bn] * 100
+                blocks.append(bn)
+                diffs.append(pct)
+        if blocks:
+            colors = ["green" if d <= 0 else "red" for d in diffs]
+            ax3.bar(blocks, diffs, width=1.0, color=colors, alpha=0.7, edgecolor="none")
+            ax3.axhline(0, color="black", linewidth=0.5)
+        ax3.set_ylabel("Δ Latency (%)")
+        ax3.set_title("Per-Block newPayload Latency Change (feature vs baseline)")
+        ax3.grid(True, alpha=0.3, axis="y")
+
+    axes[-1].set_xlabel("Block Number")
+    fig.tight_layout()
+    fig.savefig(out, dpi=150)
+    plt.close(fig)
+
+
+def plot_wait_breakdown(
+    feature: list[dict], baseline: list[dict] | None, out: Path,
+    baseline_name: str = "baseline", feature_name: str = "feature",
+):
+    series = [
+        ("Persistence Wait", "persistence_wait_us"),
+        ("State Cache Wait", "execution_cache_wait_us"),
+        ("Trie Cache Wait", "sparse_trie_wait_us"),
+    ]
+
+    fig, axes = plt.subplots(len(series), 1, figsize=(12, 3 * len(series)), sharex=True)
+    for ax, (label, key) in zip(axes, series):
+        if baseline:
+            bx = [r["block_number"] for r in baseline if r[key] is not None]
+            by = [r[key] / 1_000 for r in baseline if r[key] is not None]
+            if bx:
+                ax.plot(bx, by, linewidth=0.8, label=baseline_name, alpha=0.7)
+
+        fx = [r["block_number"] for r in feature if r[key] is not None]
+        fy = [r[key] / 1_000 for r in feature if r[key] is not None]
+        if fx:
+            ax.plot(fx, fy, linewidth=0.8, label=feature_name)
+
+        ax.set_ylabel("ms")
+        ax.set_title(label)
+        ax.grid(True, alpha=0.3)
+        if baseline:
+            ax.legend()
+
+    axes[-1].set_xlabel("Block Number")
+    fig.suptitle("Wait Time Breakdown per Block", fontsize=14, y=1.01)
+    fig.tight_layout()
+    fig.savefig(out, dpi=150, bbox_inches="tight")
+    plt.close(fig)
+
+
+def _add_regression(ax, x, y, color, label):
+    """Add a linear regression line to the axes."""
+    if len(x) < 2:
+        return
+    xa, ya = np.array(x), np.array(y)
+    m, b = np.polyfit(xa, ya, 1)
+    x_range = np.linspace(xa.min(), xa.max(), 100)
+    ax.plot(x_range, m * x_range + b, color=color, linewidth=1.5, alpha=0.8,
+            label=label)
+
+
+def plot_gas_vs_latency(
+    feature: list[dict], baseline: list[dict] | None, out: Path,
+    baseline_name: str = "baseline", feature_name: str = "feature",
+):
+    fig, ax = plt.subplots(figsize=(8, 6))
+
+    if baseline:
+        bgas = [r["gas_used"] / 1_000_000 for r in baseline]
+        blat = [r["new_payload_latency_us"] / 1_000 for r in baseline]
+        ax.scatter(bgas, blat, s=8, alpha=0.5)
+        _add_regression(ax, bgas, blat, "tab:blue", baseline_name)
+
+    fgas = [r["gas_used"] / 1_000_000 for r in feature]
+    flat = [r["new_payload_latency_us"] / 1_000 for r in feature]
+    ax.scatter(fgas, flat, s=8, alpha=0.6)
+    _add_regression(ax, fgas, flat, "tab:orange", feature_name)
+
+    ax.set_xlabel("Gas Used (Mgas)")
+    ax.set_ylabel("newPayload Latency (ms)")
+    ax.set_title("Gas Used vs Latency")
+    ax.grid(True, alpha=0.3)
+    ax.legend()
+    fig.tight_layout()
+    fig.savefig(out, dpi=150)
+    plt.close(fig)
+
+
+def merge_csvs(paths: list[str]) -> list[dict]:
+    """Parse and merge multiple CSVs, averaging values for duplicate blocks."""
+    by_block: dict[int, list[dict]] = {}
+    for path in paths:
+        for row in parse_combined_csv(path):
+            by_block.setdefault(row["block_number"], []).append(row)
+
+    merged = []
+    for bn in sorted(by_block):
+        rows = by_block[bn]
+        if len(rows) == 1:
+            merged.append(rows[0])
+        else:
+            avg = {"block_number": bn}
+            for key in ("gas_used", "new_payload_latency_us"):
+                avg[key] = int(sum(r[key] for r in rows) / len(rows))
+            for key in ("persistence_wait_us", "execution_cache_wait_us", "sparse_trie_wait_us"):
+                vals = [r[key] for r in rows if r[key] is not None]
+                avg[key] = int(sum(vals) / len(vals)) if vals else None
+            merged.append(avg)
+    return merged
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Generate benchmark charts")
+    parser.add_argument(
+        "--feature", nargs="+", required=True,
+        help="Path(s) to feature combined_latency.csv",
+    )
+    parser.add_argument(
+        "--output-dir", required=True, help="Output directory for PNG charts"
+    )
+    parser.add_argument(
+        "--baseline", nargs="+", help="Path(s) to baseline combined_latency.csv"
+    )
+    parser.add_argument("--baseline-name", default="baseline", help="Label for baseline")
+    parser.add_argument("--feature-name", "--branch-name", default="feature", help="Label for feature")
+    args = parser.parse_args()
+
+    feature = merge_csvs(args.feature)
+    if not feature:
+        print("No results found in feature CSV(s)", file=sys.stderr)
+        sys.exit(1)
+
+    baseline = None
+    if args.baseline:
+        baseline = merge_csvs(args.baseline)
+        if not baseline:
+            print(
+                "Warning: no results in baseline CSV(s), skipping comparison",
+                file=sys.stderr,
+            )
+            baseline = None
+
+    out_dir = Path(args.output_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    bname = args.baseline_name
+    fname = args.feature_name
+    plot_latency_and_throughput(feature, baseline, out_dir / "latency_throughput.png", bname, fname)
+    plot_wait_breakdown(feature, baseline, out_dir / "wait_breakdown.png", bname, fname)
+    plot_gas_vs_latency(feature, baseline, out_dir / "gas_vs_latency.png", bname, fname)
+
+    print(f"Charts written to {out_dir}")
+
+
+if __name__ == "__main__":
+    main()

.github/scripts/bench-reth-local.sh

@@ -0,0 +1,581 @@
+#!/usr/bin/env bash
+#
+# local-reth-bench.sh — Run the reth Engine API benchmark locally.
+#
+# Replicates the CI bench.yml workflow (build, snapshot, system tuning,
+# interleaved B-F-F-B execution, summary, charts) without any GitHub
+# Actions glue (no PR comments, no artifact upload, no Slack).
+#
+# Usage:
+#   local-reth-bench.sh <baseline-ref> <feature-ref> [options]
+#
+# Options:
+#   --blocks N      Number of blocks to benchmark (default: 500)
+#   --warmup N      Number of warmup blocks (default: 100)
+#   --cores N       Limit reth to N CPU cores, 0 = all available (default: 0)
+#   --samply        Enable samply profiling
+#   --tracy MODE    Tracy profiling: off, on, full (default: off)
+#   --tracy-filter F Tracy tracing filter (default: debug)
+#   --no-tune       Skip system tuning (useful on dev machines / macOS)
+#
+# Requires: the reth repo at RETH_REPO (default: ~/reth)
+#
+# Dependencies (install before first run):
+#   mc (MinIO client), schelk, cpupower, taskset, stdbuf, python3, curl,
+#   make, uv, pzstd, jq, Rust toolchain (cargo/rustup)
+#
+# The script delegates to the existing bench-reth-*.sh scripts in the reth
+# repo for the actual build, snapshot, and run steps.
+set -euo pipefail
+
+# ── PATH ──────────────────────────────────────────────────────────────
+# Ensure cargo and user-local bins (mc, uv) are visible
+export PATH="$HOME/.local/bin:$HOME/.cargo/bin:$PATH"
+
+# ── Defaults ──────────────────────────────────────────────────────────
+RETH_REPO="${RETH_REPO:-$HOME/reth}"
+BLOCKS=500
+WARMUP=100
+CORES=0
+SAMPLY=false
+TRACY="off"
+TRACY_FILTER="debug"
+TUNE=true
+BASELINE_REF=""
+FEATURE_REF=""
+
+# ── Parse arguments ──────────────────────────────────────────────────
+usage() {
+  cat <<EOF
+Usage: $(basename "$0") <baseline-ref> <feature-ref> [options]
+
+Options:
+  --blocks N         Number of blocks to benchmark (default: 500)
+  --warmup N         Number of warmup blocks (default: 100)
+  --cores N          Limit reth to N CPU cores (default: 0 = all)
+  --samply           Enable samply profiling
+  --tracy MODE       Tracy profiling: off, on, full (default: off)
+                       on   = tracing only (lower overhead)
+                       full = tracing + CPU sampling (higher overhead)
+  --tracy-filter F   Tracy tracing filter (default: debug)
+  --no-tune          Skip system tuning
+EOF
+  exit 1
+}
+
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --blocks)       BLOCKS="$2"; shift 2 ;;
+    --warmup)       WARMUP="$2"; shift 2 ;;
+    --cores)        CORES="$2"; shift 2 ;;
+    --samply)       SAMPLY=true; shift ;;
+    --tracy)        TRACY="$2"; shift 2 ;;
+    --tracy-filter) TRACY_FILTER="$2"; shift 2 ;;
+    --no-tune)      TUNE=false; shift ;;
+    --help|-h)  usage ;;
+    -*)         echo "Unknown option: $1"; usage ;;
+    *)
+      if [ -z "$BASELINE_REF" ]; then
+        BASELINE_REF="$1"
+      elif [ -z "$FEATURE_REF" ]; then
+        FEATURE_REF="$1"
+      else
+        echo "Unexpected argument: $1"; usage
+      fi
+      shift
+      ;;
+  esac
+done
+
+if [ -z "$BASELINE_REF" ] || [ -z "$FEATURE_REF" ]; then
+  echo "Error: both <baseline-ref> and <feature-ref> are required."
+  usage
+fi
+
+# Validate --tracy value
+case "$TRACY" in
+  off|on|full) ;;
+  *) echo "Error: --tracy must be off, on, or full (got: $TRACY)"; usage ;;
+esac
+
+# Samply + tracy=full are mutually exclusive (both use perf sampling)
+if [ "$SAMPLY" = "true" ] && [ "$TRACY" = "full" ]; then
+  echo "Warning: samply and tracy=full both use perf sampling; downgrading tracy to 'on'."
+  TRACY="on"
+fi
+
+# ── Check dependencies ───────────────────────────────────────────────
+missing=()
+for cmd in mc schelk cpupower taskset stdbuf python3 curl make uv pzstd jq cargo; do
+  command -v "$cmd" &>/dev/null || missing+=("$cmd")
+done
+if [ ${#missing[@]} -gt 0 ]; then
+  echo "Error: missing required tools: ${missing[*]}"
+  echo "See the CI 'Install dependencies' step in .github/workflows/bench.yml for install instructions."
+  exit 1
+fi
+
+if [ "$TRACY" != "off" ]; then
+  if ! command -v tracy-capture &>/dev/null; then
+    echo "Error: tracy-capture is required for --tracy $TRACY"
+    exit 1
+  fi
+fi
+
+# Ensure tools that run via sudo are in a sudo-visible path.
+# The bench scripts use `sudo schelk` / `sudo samply` but cargo installs
+# them to ~/.cargo/bin which sudo's secure_path doesn't include.
+for cmd in schelk samply; do
+  if command -v "$cmd" &>/dev/null && ! sudo sh -c "command -v $cmd" &>/dev/null; then
+    echo "Installing $cmd to /usr/local/bin (needed for sudo)..."
+    sudo install "$(command -v "$cmd")" /usr/local/bin/
+  fi
+done
+
+if [ ! -d "$RETH_REPO/.git" ]; then
+  echo "Error: RETH_REPO=$RETH_REPO is not a git repository."
+  echo "Set RETH_REPO or clone reth to ~/reth"
+  exit 1
+fi
+
+# ── Resolve paths ────────────────────────────────────────────────────
+SELF_DIR="$(cd "$(dirname "$0")" && pwd)"
+SCRIPTS_DIR="${RETH_REPO}/.github/scripts"
+BENCH_WORK_DIR="${RETH_REPO}/../bench-work-$(date +%Y%m%d-%H%M%S)"
+BASELINE_SRC="${RETH_REPO}/../reth-baseline"
+FEATURE_SRC="${RETH_REPO}/../reth-feature"
+
+mkdir -p "$BENCH_WORK_DIR"
+BENCH_WORK_DIR="$(cd "$BENCH_WORK_DIR" && pwd)"
+
+# ── Global cleanup trap (restores system tuning on any exit) ─────────
+TUNING_APPLIED=false
+CSTATE_PID=
+METRICS_PROXY_PID=
+cleanup_global() {
+  [ -n "$METRICS_PROXY_PID" ] && kill "$METRICS_PROXY_PID" 2>/dev/null || true
+  if [ "$TUNING_APPLIED" = true ]; then
+    echo
+    echo "▸ Restoring system settings..."
+    [ -n "$CSTATE_PID" ] && kill "$CSTATE_PID" 2>/dev/null || true
+    sudo systemctl start irqbalance cron atd 2>/dev/null || true
+    echo "  System settings restored."
+  fi
+}
+trap cleanup_global EXIT
+
+echo "═══════════════════════════════════════════════════════════"
+echo "  reth local benchmark"
+echo "═══════════════════════════════════════════════════════════"
+echo "  Baseline ref : $BASELINE_REF"
+echo "  Feature ref  : $FEATURE_REF"
+echo "  Blocks       : $BLOCKS"
+echo "  Warmup       : $WARMUP"
+echo "  Cores        : $CORES"
+echo "  Samply       : $SAMPLY"
+echo "  Tracy        : $TRACY"
+echo "  Tracy filter : $TRACY_FILTER"
+echo "  System tune  : $TUNE"
+echo "  Work dir     : $BENCH_WORK_DIR"
+echo "  Reth repo    : $RETH_REPO"
+echo "═══════════════════════════════════════════════════════════"
+echo
+
+# Enable sccache if available (matches CI's RUSTC_WRAPPER=sccache)
+if command -v sccache &>/dev/null; then
+  export RUSTC_WRAPPER="sccache"
+fi
+
+# Export env vars expected by the bench-reth-*.sh scripts
+export BENCH_BLOCKS="$BLOCKS"
+export BENCH_WARMUP_BLOCKS="$WARMUP"
+export BENCH_CORES="$CORES"
+export BENCH_SAMPLY="$SAMPLY"
+export BENCH_TRACY="$TRACY"
+export BENCH_TRACY_FILTER="$TRACY_FILTER"
+export BENCH_WORK_DIR
+export SCHELK_MOUNT="${SCHELK_MOUNT:-/reth-bench}"
+export BENCH_RPC_URL="${BENCH_RPC_URL:-https://ethereum.reth.rs/rpc}"
+export BENCH_METRICS_ADDR="127.0.0.1:9100"
+
+# ── Step 1: Resolve refs to full SHAs ────────────────────────────────
+echo "▸ Resolving git refs..."
+cd "$RETH_REPO"
+
+resolve_ref() {
+  local ref="$1"
+  git fetch origin "$ref" --quiet 2>/dev/null || true
+  git rev-parse "$ref" 2>/dev/null \
+    || git rev-parse "origin/$ref" 2>/dev/null \
+    || { echo "Error: cannot resolve ref '$ref'"; exit 1; }
+}
+
+BASELINE_SHA="$(resolve_ref "$BASELINE_REF")"
+FEATURE_SHA="$(resolve_ref "$FEATURE_REF")"
+echo "  Baseline SHA : $BASELINE_SHA"
+echo "  Feature SHA  : $FEATURE_SHA"
+echo
+
+# ── Step 2: Prepare source directories ───────────────────────────────
+echo "▸ Preparing source directories..."
+
+prepare_source() {
+  local src_dir="$1" ref="$2"
+  if [ -d "$src_dir" ]; then
+    git -C "$src_dir" fetch origin "$ref" 2>/dev/null || true
+  else
+    git clone --recurse-submodules "$RETH_REPO" "$src_dir"
+  fi
+  git -C "$src_dir" checkout "$ref" --force
+  git -C "$src_dir" submodule update --init --recursive
+}
+
+prepare_source "$BASELINE_SRC" "$BASELINE_SHA"
+prepare_source "$FEATURE_SRC" "$FEATURE_SHA"
+BASELINE_SRC="$(cd "$BASELINE_SRC" && pwd)"
+FEATURE_SRC="$(cd "$FEATURE_SRC" && pwd)"
+echo "  Baseline src : $BASELINE_SRC"
+echo "  Feature src  : $FEATURE_SRC"
+echo
+
+# ── Step 3: Check / download snapshot ────────────────────────────────
+echo "▸ Checking snapshot..."
+cd "$RETH_REPO"
+SNAPSHOT_NEEDED=false
+if ! "${SCRIPTS_DIR}/bench-reth-snapshot.sh" --check; then
+  SNAPSHOT_NEEDED=true
+  echo "  Snapshot needs update."
+else
+  echo "  Snapshot is up-to-date."
+fi
+echo
+
+# ── Step 4: Build binaries (+ snapshot download) in parallel ─────────
+echo "▸ Building binaries (parallel)..."
+cd "$RETH_REPO"
+
+FAIL=0
+
+"${SCRIPTS_DIR}/bench-reth-build.sh" baseline "$BASELINE_SRC" "$BASELINE_SHA" &
+PID_BASELINE=$!
+
+"${SCRIPTS_DIR}/bench-reth-build.sh" feature "$FEATURE_SRC" "$FEATURE_SHA" &
+PID_FEATURE=$!
+
+PID_SNAPSHOT=
+if [ "$SNAPSHOT_NEEDED" = "true" ]; then
+  echo "  Also downloading snapshot in parallel..."
+  "${SCRIPTS_DIR}/bench-reth-snapshot.sh" &
+  PID_SNAPSHOT=$!
+fi
+
+wait $PID_BASELINE || FAIL=1
+wait $PID_FEATURE  || FAIL=1
+[ -n "$PID_SNAPSHOT" ] && { wait $PID_SNAPSHOT || FAIL=1; }
+
+if [ $FAIL -ne 0 ]; then
+  echo "Error: one or more parallel tasks failed (builds / snapshot)"
+  exit 1
+fi
+echo "  Binaries built successfully."
+echo
+
+# ── Step 5: System tuning (optional) ────────────────────────────────
+if [ "$TUNE" = "true" ]; then
+  echo "▸ Applying system tuning..."
+
+  sudo cpupower frequency-set -g performance 2>/dev/null || true
+
+  # Disable turbo boost (Intel + AMD)
+  echo 1 | sudo tee /sys/devices/system/cpu/intel_pstate/no_turbo 2>/dev/null || true
+  echo 0 | sudo tee /sys/devices/system/cpu/cpufreq/boost 2>/dev/null || true
+
+  sudo swapoff -a 2>/dev/null || true
+  echo 0 | sudo tee /proc/sys/kernel/randomize_va_space 2>/dev/null || true
+
+  # Disable SMT (hyperthreading)
+  for cpu in /sys/devices/system/cpu/cpu*/topology/thread_siblings_list; do
+    [ -f "$cpu" ] || continue
+    first=$(cut -d, -f1 < "$cpu" | cut -d- -f1)
+    current=$(echo "$cpu" | grep -o 'cpu[0-9]*' | grep -o '[0-9]*')
+    if [ "$current" != "$first" ]; then
+      echo 0 | sudo tee "/sys/devices/system/cpu/cpu${current}/online" 2>/dev/null || true
+    fi
+  done
+  echo "  Online CPUs: $(nproc)"
+
+  # Disable transparent huge pages
+  for p in /sys/kernel/mm/transparent_hugepage /sys/kernel/mm/transparent_hugepages; do
+    if [ -d "$p" ]; then
+      echo never | sudo tee "$p/enabled" 2>/dev/null || true
+      echo never | sudo tee "$p/defrag" 2>/dev/null || true
+      break
+    fi
+  done
+
+  # Prevent deep C-states
+  sudo sh -c 'exec 3<>/dev/cpu_dma_latency; echo -ne "\x00\x00\x00\x00" >&3; sleep infinity' &
+  CSTATE_PID=$!
+
+  # Pin IRQs to core 0
+  for irq in /proc/irq/*/smp_affinity_list; do
+    echo 0 | sudo tee "$irq" 2>/dev/null || true
+  done
+
+  # Stop noisy background services
+  sudo systemctl stop irqbalance cron atd unattended-upgrades snapd 2>/dev/null || true
+
+  TUNING_APPLIED=true
+
+  # Log environment for reproducibility (matches CI)
+  echo "  === Benchmark environment ==="
+  echo "  Kernel : $(uname -r)"
+  lscpu | grep -E 'Model name|CPU\(s\)|MHz|NUMA' | sed 's/^/  /'
+  echo "  Governor : $(cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor 2>/dev/null || echo unknown)"
+  echo "  Freq     : $(cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq 2>/dev/null || echo unknown)"
+  echo "  THP      : $(cat /sys/kernel/mm/transparent_hugepage/enabled 2>/dev/null || cat /sys/kernel/mm/transparent_hugepages/enabled 2>/dev/null || echo unknown)"
+  free -h | sed 's/^/  /'
+  echo "  System tuning applied."
+  echo
+fi
+
+# ── Step 5b: Tracefs mount (tracy=full only) ─────────────────────────
+if [ "$TRACY" = "full" ] && [ "$(uname)" = "Linux" ]; then
+  echo "▸ Mounting tracefs for Tracy full mode..."
+  sudo mount -t tracefs tracefs /sys/kernel/tracing -o mode=755 2>/dev/null || true
+fi
+
+# ── Tracy upload & viewer helpers ────────────────────────────────────
+TRACY_VIEWER_BASE="${TRACY_VIEWER_BASE:-}"
+
+tracy_viewer_url() {
+  local profile_url="$1"
+  if [ -z "$TRACY_VIEWER_BASE" ]; then
+    echo ""
+    return
+  fi
+  local encoded
+  encoded=$(python3 -c "import urllib.parse, sys; print(urllib.parse.quote(sys.argv[1], safe=''))" "$profile_url")
+  echo "${TRACY_VIEWER_BASE}?profile_url=${encoded}"
+}
+
+upload_tracy() {
+  local label="$1" output_dir="$2" sha="$3"
+  local tracy_file="$output_dir/tracy-profile.tracy"
+
+  if [ ! -f "$tracy_file" ]; then
+    echo "  Tracy: no profile found, skipping upload."
+    return
+  fi
+
+  local timestamp short_sha remote_name bucket mc_alias
+  timestamp=$(date +%Y%m%d-%H%M%S)
+  short_sha="${sha:0:7}"
+  remote_name="${label}-${short_sha}-${timestamp}.tracy"
+  bucket="${TRACY_BUCKET:-tracy-profiles}"
+  mc_alias="${MC_ALIAS:-minio}"
+  local minio_base="${TRACY_MINIO_URL:-http://minio.minio.svc.cluster.local:9000}"
+
+  echo "  Tracy: uploading profile..."
+  if mc cp "$tracy_file" "${mc_alias}/${bucket}/${remote_name}"; then
+    local url="${minio_base}/${bucket}/${remote_name}"
+    echo "$url" > "$output_dir/tracy_url.txt"
+    local viewer
+    viewer=$(tracy_viewer_url "$url")
+    if [ -n "$viewer" ]; then
+      echo "$viewer" > "$output_dir/tracy_viewer_url.txt"
+      echo "  Tracy: uploaded → $viewer"
+    else
+      echo "  Tracy: uploaded → $url"
+    fi
+  else
+    echo "  Tracy: upload failed (non-fatal)."
+  fi
+
+  # Delete large profile to free disk
+  rm -f "$tracy_file"
+}
+
+# ── Step 6: Pre-flight cleanup ───────────────────────────────────────
+echo "▸ Pre-flight cleanup..."
+pkill -f bench-metrics-proxy 2>/dev/null || true
+sudo pkill -9 reth 2>/dev/null || true
+sleep 1
+if mountpoint -q "$SCHELK_MOUNT" 2>/dev/null; then
+  sudo umount -l "$SCHELK_MOUNT" 2>/dev/null || true
+  sudo schelk recover -y 2>/dev/null || true
+fi
+echo
+
+# ── Step 7: Interleaved benchmark runs (B-F-F-B) ────────────────────
+# This ordering reduces systematic bias from thermal drift and cache warming.
+BASELINE_BIN="${BASELINE_SRC}/target/profiling/reth"
+FEATURE_BIN="${FEATURE_SRC}/target/profiling/reth"
+
+# Start metrics proxy (reth → label injection → Prometheus)
+LABELS_FILE="/tmp/bench-metrics-labels.json"
+echo '{}' > "$LABELS_FILE"
+METRICS_SUBNET="${METRICS_SUBNET:-10.10.0.0/24}"
+METRICS_PORT="${METRICS_PORT:-9090}"
+python3 "${SELF_DIR}/bench-metrics-proxy.py" \
+  --labels "$LABELS_FILE" \
+  --upstream "http://${BENCH_METRICS_ADDR}/" \
+  --subnet "$METRICS_SUBNET" \
+  --port "$METRICS_PORT" &
+METRICS_PROXY_PID=$!
+echo "▸ Metrics proxy started (PID $METRICS_PROXY_PID) on subnet ${METRICS_SUBNET}, port ${METRICS_PORT}"
+
+# Unique benchmark ID: local-<timestamp> for local runs, ci-<run_id> for CI
+BENCH_ID="local-$(basename "$BENCH_WORK_DIR" | sed 's/bench-work-//')"
+# Reference epoch: shared time origin so all runs overlay in Grafana.
+# The proxy maps each run's elapsed time onto this common origin.
+BENCH_REFERENCE_EPOCH=$(date +%s)
+
+write_labels() {
+  local run_label="$1" run_type="$2" ref="$3" sha="$4"
+  LAST_RUN_START=$(date +%s)
+  cat > "$LABELS_FILE" <<-EOF
+	{"benchmark_run":"${run_label}","run_type":"${run_type}","git_ref":"${ref}","bench_sha":"${sha}","benchmark_id":"${BENCH_ID}","run_start_epoch":"${LAST_RUN_START}","reference_epoch":"${BENCH_REFERENCE_EPOCH}"}
+	EOF
+}
+
+run_bench() {
+  local label="$1" binary="$2" output_dir="$3"
+  echo "▸ Running benchmark: ${label}..."
+  cd "$RETH_REPO"
+  if command -v taskset &>/dev/null; then
+    taskset -c 0 "${SCRIPTS_DIR}/bench-reth-run.sh" "$label" "$binary" "$output_dir"
+  else
+    "${SCRIPTS_DIR}/bench-reth-run.sh" "$label" "$binary" "$output_dir"
+  fi
+  echo "  ✓ ${label} complete."
+  echo
+}
+
+write_labels "baseline-1" "baseline" "$BASELINE_REF" "$BASELINE_SHA"
+run_bench "baseline-1" "$BASELINE_BIN" "$BENCH_WORK_DIR/baseline-1"
+
+write_labels "feature-1" "feature" "$FEATURE_REF" "$FEATURE_SHA"
+run_bench "feature-1"  "$FEATURE_BIN"  "$BENCH_WORK_DIR/feature-1"
+
+write_labels "feature-2" "feature" "$FEATURE_REF" "$FEATURE_SHA"
+run_bench "feature-2"  "$FEATURE_BIN"  "$BENCH_WORK_DIR/feature-2"
+
+write_labels "baseline-2" "baseline" "$BASELINE_REF" "$BASELINE_SHA"
+run_bench "baseline-2" "$BASELINE_BIN" "$BENCH_WORK_DIR/baseline-2"
+
+# ── Compute Grafana URL ──────────────────────────────────────────────
+GRAFANA_BASE_URL="https://tempoxyz.grafana.net/d/reth-bench-ghr/reth-bench-ghr"
+GRAFANA_DATASOURCE="ef57fux92e9z4e"
+LAST_RUN_DURATION=$(( $(date +%s) - LAST_RUN_START ))
+FROM_MS=$(( BENCH_REFERENCE_EPOCH * 1000 ))
+TO_MS=$(( (BENCH_REFERENCE_EPOCH + LAST_RUN_DURATION) * 1000 ))
+GRAFANA_URL="${GRAFANA_BASE_URL}?orgId=1&from=${FROM_MS}&to=${TO_MS}&timezone=browser&var-datasource=${GRAFANA_DATASOURCE}&var-job=reth-bench&var-benchmark_id=${BENCH_ID}&var-benchmark_run=\$__all"
+
+# ── Step 8: Scan logs for errors ─────────────────────────────────────
+echo "▸ Scanning logs for errors..."
+ERRORS_FILE="$BENCH_WORK_DIR/errors.md"
+found_errors=false
+for run_dir in baseline-1 feature-1 feature-2 baseline-2; do
+  LOG="$BENCH_WORK_DIR/$run_dir/node.log"
+  [ -f "$LOG" ] || continue
+  panics=$(grep -c -E 'panicked at' "$LOG" 2>/dev/null || true)
+  errors=$(grep -c ' ERROR ' "$LOG" 2>/dev/null || true)
+  if [ "$panics" -gt 0 ] || [ "$errors" -gt 0 ]; then
+    if [ "$found_errors" = false ]; then
+      printf '### ⚠️ Node Errors\n\n' >> "$ERRORS_FILE"
+      found_errors=true
+    fi
+    printf '<details><summary><b>%s</b>: %d panic(s), %d error(s)</summary>\n\n' \
+      "$run_dir" "$panics" "$errors" >> "$ERRORS_FILE"
+    if [ "$panics" -gt 0 ]; then
+      printf '**Panics:**\n```\n' >> "$ERRORS_FILE"
+      grep -E 'panicked at' "$LOG" | head -10 >> "$ERRORS_FILE"
+      printf '```\n' >> "$ERRORS_FILE"
+    fi
+    if [ "$errors" -gt 0 ]; then
+      printf '**Errors (first 20):**\n```\n' >> "$ERRORS_FILE"
+      grep ' ERROR ' "$LOG" | head -20 >> "$ERRORS_FILE"
+      printf '```\n' >> "$ERRORS_FILE"
+    fi
+    printf '\n</details>\n\n' >> "$ERRORS_FILE"
+  fi
+done
+if [ "$found_errors" = true ]; then
+  echo "  ⚠ Errors found — see $ERRORS_FILE"
+else
+  echo "  No errors found."
+fi
+echo
+
+# ── Step 9: Parse results ───────────────────────────────────────────
+echo "▸ Parsing results..."
+cd "$RETH_REPO"
+
+SUMMARY_ARGS=(
+  --output-summary "$BENCH_WORK_DIR/summary.json"
+  --output-markdown "$BENCH_WORK_DIR/comment.md"
+  --repo "paradigmxyz/reth"
+  --baseline-ref "$BASELINE_SHA"
+  --baseline-name "$BASELINE_REF"
+  --feature-name "$FEATURE_REF"
+  --feature-ref "$FEATURE_SHA"
+  --baseline-csv "$BENCH_WORK_DIR/baseline-1/combined_latency.csv" "$BENCH_WORK_DIR/baseline-2/combined_latency.csv"
+  --feature-csv "$BENCH_WORK_DIR/feature-1/combined_latency.csv" "$BENCH_WORK_DIR/feature-2/combined_latency.csv"
+  --gas-csv "$BENCH_WORK_DIR/feature-1/total_gas.csv"
+  --grafana-url "$GRAFANA_URL"
+)
+
+python3 "${SCRIPTS_DIR}/bench-reth-summary.py" "${SUMMARY_ARGS[@]}"
+echo
+
+# ── Step 10: Generate charts ─────────────────────────────────────────
+echo "▸ Generating charts..."
+CHART_ARGS=(
+  --output-dir "$BENCH_WORK_DIR/charts"
+  --feature "$BENCH_WORK_DIR/feature-1/combined_latency.csv" "$BENCH_WORK_DIR/feature-2/combined_latency.csv"
+  --baseline "$BENCH_WORK_DIR/baseline-1/combined_latency.csv" "$BENCH_WORK_DIR/baseline-2/combined_latency.csv"
+  --baseline-name "$BASELINE_REF"
+  --feature-name "$FEATURE_REF"
+)
+
+if python3 -c "import matplotlib" 2>/dev/null; then
+  python3 "${SCRIPTS_DIR}/bench-reth-charts.py" "${CHART_ARGS[@]}"
+elif command -v uv &>/dev/null; then
+  uv run --with matplotlib python3 "${SCRIPTS_DIR}/bench-reth-charts.py" "${CHART_ARGS[@]}"
+else
+  echo "  Warning: matplotlib not available, skipping chart generation."
+fi
+echo
+
+# ── Step 11: Upload Tracy profiles ────────────────────────────────────
+if [ "$TRACY" != "off" ]; then
+  echo "▸ Uploading Tracy profiles..."
+  upload_tracy "baseline-1" "$BENCH_WORK_DIR/baseline-1" "$BASELINE_SHA"
+  upload_tracy "feature-1"  "$BENCH_WORK_DIR/feature-1"  "$FEATURE_SHA"
+  upload_tracy "feature-2"  "$BENCH_WORK_DIR/feature-2"  "$FEATURE_SHA"
+  upload_tracy "baseline-2" "$BENCH_WORK_DIR/baseline-2" "$BASELINE_SHA"
+  echo
+fi
+
+# ── Done (system restore happens via EXIT trap) ─────────────────────
+echo "═══════════════════════════════════════════════════════════"
+echo "  Benchmark complete!"
+echo "═══════════════════════════════════════════════════════════"
+echo "  Results  : $BENCH_WORK_DIR/summary.json"
+echo "  Markdown : $BENCH_WORK_DIR/comment.md"
+echo "  Charts   : $BENCH_WORK_DIR/charts/"
+if [ -f "$ERRORS_FILE" ]; then
+  echo "  Errors   : $ERRORS_FILE"
+fi
+echo "  Grafana  : $GRAFANA_URL"
+if [ "$TRACY" != "off" ]; then
+  echo "  ─── Tracy Profiles ───"
+  for run_dir in baseline-1 feature-1 feature-2 baseline-2; do
+    url_file="$BENCH_WORK_DIR/$run_dir/tracy_viewer_url.txt"
+    if [ -f "$url_file" ]; then
+      echo "  $run_dir : $(cat "$url_file")"
+    fi
+  done
+fi
+echo "═══════════════════════════════════════════════════════════"

.github/scripts/bench-reth-run.sh

@@ -0,0 +1,307 @@
+#!/usr/bin/env bash
+#
+# Runs a single reth-bench cycle: mount snapshot → start node → warmup →
+# benchmark → stop node → recover snapshot.
+#
+# Usage: bench-reth-run.sh <label> <binary> <output-dir>
+#
+# Required env: SCHELK_MOUNT, BENCH_RPC_URL, BENCH_BLOCKS, BENCH_WARMUP_BLOCKS
+# Optional env: BENCH_BIG_BLOCKS (true/false), BENCH_WORK_DIR (for big blocks path)
+#               BENCH_RETH_NEW_PAYLOAD (true/false, default true)
+#               BENCH_WAIT_TIME (duration like 500ms, default empty)
+#               BENCH_BASELINE_ARGS (extra reth node args for baseline runs)
+#               BENCH_FEATURE_ARGS (extra reth node args for feature runs)
+#               BENCH_OTLP_TRACES_ENDPOINT (OTLP HTTP endpoint for traces, e.g. https://host/insert/opentelemetry/v1/traces)
+#               BENCH_OTLP_LOGS_ENDPOINT (OTLP HTTP endpoint for logs, e.g. https://host/insert/opentelemetry/v1/logs)
+set -euo pipefail
+
+LABEL="$1"
+BINARY="$2"
+OUTPUT_DIR="$3"
+DATADIR_NAME="datadir"
+if [ "${BENCH_BIG_BLOCKS:-false}" = "true" ]; then
+  DATADIR_NAME="datadir-big-blocks"
+fi
+DATADIR="$SCHELK_MOUNT/$DATADIR_NAME"
+mkdir -p "$OUTPUT_DIR"
+LOG="${OUTPUT_DIR}/node.log"
+
+cleanup() {
+  kill "$TAIL_PID" 2>/dev/null || true
+  # Stop tracy-capture first (SIGINT makes it disconnect and flush to disk)
+  # Must happen before killing reth, otherwise reth keeps streaming data.
+  if [ -n "${TRACY_PID:-}" ] && kill -0 "$TRACY_PID" 2>/dev/null; then
+    echo "Stopping tracy-capture..."
+    kill -INT "$TRACY_PID" 2>/dev/null || true
+    for i in $(seq 1 30); do
+      kill -0 "$TRACY_PID" 2>/dev/null || break
+      if [ $((i % 10)) -eq 0 ]; then
+        echo "Waiting for tracy-capture to finish writing... (${i}s)"
+      fi
+      sleep 1
+    done
+    if kill -0 "$TRACY_PID" 2>/dev/null; then
+      echo "tracy-capture still running after 30s, killing..."
+      kill -9 "$TRACY_PID" 2>/dev/null || true
+    fi
+    wait "$TRACY_PID" 2>/dev/null || true
+  fi
+  if [ -n "${RETH_PID:-}" ] && sudo kill -0 "$RETH_PID" 2>/dev/null; then
+    if [ "${BENCH_SAMPLY:-false}" = "true" ]; then
+      # Send SIGINT to the inner reth process by exact name (not -f which
+      # would also match samply's cmdline containing "reth"). Samply will
+      # capture reth's exit and save the profile.
+      sudo pkill -INT -x reth 2>/dev/null || true
+      # Wait for samply to finish writing the profile and exit
+      for i in $(seq 1 120); do
+        sudo pgrep -x samply > /dev/null 2>&1 || break
+        if [ $((i % 10)) -eq 0 ]; then
+          echo "Waiting for samply to finish writing profile... (${i}s)"
+        fi
+        sleep 1
+      done
+      if sudo pgrep -x samply > /dev/null 2>&1; then
+        echo "Samply still running after 120s, sending SIGTERM..."
+        sudo pkill -x samply 2>/dev/null || true
+      fi
+    else
+      sudo kill "$RETH_PID"
+      for i in $(seq 1 30); do
+        sudo kill -0 "$RETH_PID" 2>/dev/null || break
+        sleep 1
+      done
+    fi
+    sudo kill -9 "$RETH_PID" 2>/dev/null || true
+    sleep 1
+  fi
+  # Fix ownership of reth-created files (reth runs as root)
+  sudo chown -R "$(id -un):$(id -gn)" "$OUTPUT_DIR" 2>/dev/null || true
+  if mountpoint -q "$SCHELK_MOUNT"; then
+    sudo umount -l "$SCHELK_MOUNT" || true
+    sudo schelk recover -y || true
+  fi
+}
+TAIL_PID=
+TRACY_PID=
+trap cleanup EXIT
+
+# Clean up stale schelk state from a previous cancelled run.
+# If schelk thinks it's still mounted (e.g. a cancelled run skipped cleanup),
+# recover first to reset state.
+sudo schelk recover -y -k || true
+
+# Mount
+sudo schelk mount -y
+sync
+sudo sh -c 'echo 3 > /proc/sys/vm/drop_caches'
+echo "=== Cache state after drop ==="
+free -h
+grep Cached /proc/meminfo
+
+# Start reth
+# CPU layout: core 0 = OS/IRQs/reth-bench/aux, cores 1+ = reth node
+RETH_BENCH="$(which reth-bench)"
+ONLINE=$(nproc --all)
+MAX_RETH=$(( ONLINE - 1 ))
+if [ "${BENCH_CORES:-0}" -gt 0 ] && [ "$BENCH_CORES" -lt "$MAX_RETH" ]; then
+  MAX_RETH=$BENCH_CORES
+fi
+RETH_CPUS="1-${MAX_RETH}"
+
+BIG_BLOCKS="${BENCH_BIG_BLOCKS:-false}"
+
+RETH_ARGS=(
+  node
+  --datadir "$DATADIR"
+  --log.file.directory "$OUTPUT_DIR/reth-logs"
+  --engine.accept-execution-requests-hash
+  --http
+  --http.port 8545
+  --ws
+  --ws.api all
+  --authrpc.port 8551
+  --disable-discovery
+  --no-persist-peers
+)
+
+# Gate flag on binary support (older baselines may not have it).
+# Uses --help which exits immediately via clap without node init.
+SYNC_STATE_IDLE=false
+if "$BINARY" node --help 2>/dev/null | grep -qF -- '--debug.startup-sync-state-idle'; then
+  RETH_ARGS+=(--debug.startup-sync-state-idle)
+  SYNC_STATE_IDLE=true
+fi
+
+# Append per-label extra node args (baseline or feature)
+EXTRA_NODE_ARGS=""
+case "$LABEL" in
+  baseline*) EXTRA_NODE_ARGS="${BENCH_BASELINE_ARGS:-}" ;;
+  feature*)  EXTRA_NODE_ARGS="${BENCH_FEATURE_ARGS:-}" ;;
+esac
+if [ -n "$EXTRA_NODE_ARGS" ]; then
+  # Word-split the string into individual args
+  # shellcheck disable=SC2206
+  RETH_ARGS+=($EXTRA_NODE_ARGS)
+fi
+
+if [ -n "${BENCH_METRICS_ADDR:-}" ]; then
+  RETH_ARGS+=(--metrics "$BENCH_METRICS_ADDR")
+fi
+
+# OTLP traces and logs export
+if [ -n "${BENCH_OTLP_TRACES_ENDPOINT:-}" ]; then
+  RETH_ARGS+=(--tracing-otlp="${BENCH_OTLP_TRACES_ENDPOINT}" --tracing-otlp.service-name=reth-bench)
+fi
+if [ -n "${BENCH_OTLP_LOGS_ENDPOINT:-}" ]; then
+  RETH_ARGS+=(--logs-otlp="${BENCH_OTLP_LOGS_ENDPOINT}" --logs-otlp.filter=debug)
+fi
+
+# Tracy profiling: add --log.tracy flags and set environment
+if [ "${BENCH_TRACY:-off}" != "off" ]; then
+  RETH_ARGS+=(--log.tracy --log.tracy.filter "${BENCH_TRACY_FILTER:-debug}")
+  if [ "${BENCH_TRACY}" = "on" ]; then
+    export TRACY_NO_SYS_TRACE=1
+  elif [ "${BENCH_TRACY}" = "full" ]; then
+    export TRACY_SAMPLING_HZ="${BENCH_TRACY_SAMPLING_HZ:-1}"
+  fi
+fi
+
+SUDO_ENV=()
+if [ -n "${OTEL_RESOURCE_ATTRIBUTES:-}" ]; then
+  SUDO_ENV+=("OTEL_RESOURCE_ATTRIBUTES=${OTEL_RESOURCE_ATTRIBUTES}")
+  SUDO_ENV+=("OTEL_BSP_MAX_QUEUE_SIZE=65536" "OTEL_BLRP_MAX_QUEUE_SIZE=65536")
+fi
+
+# Limit reth memory to 95% of available RAM to prevent OOM kills
+TOTAL_MEM_KB=$(awk '/^MemTotal:/ {print $2}' /proc/meminfo)
+MEM_LIMIT=$(( TOTAL_MEM_KB * 95 / 100 * 1024 ))
+echo "Memory limit: $(( MEM_LIMIT / 1024 / 1024 ))MB (95% of $(( TOTAL_MEM_KB / 1024 ))MB)"
+
+if [ "${BENCH_SAMPLY:-false}" = "true" ]; then
+  RETH_ARGS+=(--log.samply)
+  SAMPLY="$(which samply)"
+  sudo systemd-run --scope -p MemoryMax="$MEM_LIMIT" -p AllowedCPUs="$RETH_CPUS" \
+    env "${SUDO_ENV[@]}" nice -n -20 \
+    "$SAMPLY" record --save-only --presymbolicate --rate 10000 \
+    --output "$OUTPUT_DIR/samply-profile.json.gz" \
+    -- "$BINARY" "${RETH_ARGS[@]}" \
+    > "$LOG" 2>&1 &
+else
+  sudo systemd-run --scope -p MemoryMax="$MEM_LIMIT" -p AllowedCPUs="$RETH_CPUS" \
+    env "${SUDO_ENV[@]}" nice -n -20 "$BINARY" "${RETH_ARGS[@]}" \
+    > "$LOG" 2>&1 &
+fi
+
+RETH_PID=$!
+stdbuf -oL tail -f "$LOG" | sed -u "s/^/[reth] /" &
+TAIL_PID=$!
+
+for i in $(seq 1 60); do
+  if curl -sf http://127.0.0.1:8545 -X POST \
+    -H 'Content-Type: application/json' \
+    -d '{"jsonrpc":"2.0","method":"eth_blockNumber","params":[],"id":1}' \
+    > /dev/null 2>&1; then
+    echo "reth (${LABEL}) RPC is up after ${i}s"
+    break
+  fi
+  if [ "$i" -eq 60 ]; then
+    echo "::error::reth (${LABEL}) failed to start within 60s"
+    cat "$LOG"
+    exit 1
+  fi
+  sleep 1
+done
+
+# Wait for the pipeline to finish (eth_syncing returns false) so the
+# engine is in live mode and can accept newPayload calls.
+# Only possible when --debug.startup-sync-state-idle is supported.
+if [ "$SYNC_STATE_IDLE" = "true" ]; then
+  for i in $(seq 1 300); do
+    SYNC_RESULT=$(curl -sf http://127.0.0.1:8545 -X POST \
+      -H 'Content-Type: application/json' \
+      -d '{"jsonrpc":"2.0","method":"eth_syncing","params":[],"id":1}' 2>/dev/null || true)
+    if [ -n "$SYNC_RESULT" ] && jq -e '.result == false' <<< "$SYNC_RESULT" > /dev/null 2>&1; then
+      echo "reth (${LABEL}) pipeline finished after ${i}s, engine is live"
+      break
+    fi
+    if [ "$i" -eq 300 ]; then
+      echo "::error::reth (${LABEL}) pipeline did not finish within 300s"
+      cat "$LOG"
+      exit 1
+    fi
+    sleep 1
+  done
+else
+  echo "reth (${LABEL}) binary does not support --debug.startup-sync-state-idle, skipping sync wait"
+fi
+
+# Run reth-bench with high priority but as the current user so output
+# files are not root-owned (avoids EACCES on next checkout).
+BENCH_NICE="sudo nice -n -20 sudo -u $(id -un)"
+
+# Build optional flags
+EXTRA_BENCH_ARGS=()
+if [ "${BENCH_RETH_NEW_PAYLOAD:-true}" != "false" ]; then
+  EXTRA_BENCH_ARGS+=(--reth-new-payload --wait-for-persistence)
+fi
+if [ -n "${BENCH_WAIT_TIME:-}" ]; then
+  EXTRA_BENCH_ARGS+=(--wait-time "$BENCH_WAIT_TIME")
+fi
+
+if [ "$BIG_BLOCKS" = "true" ]; then
+  # Big blocks mode: replay pre-generated payloads
+  BIG_BLOCKS_DIR="${BENCH_BIG_BLOCKS_DIR:-${BENCH_WORK_DIR}/big-blocks}"
+
+  # Start tracy-capture so profile only covers the benchmark
+  if [ "${BENCH_TRACY:-off}" != "off" ]; then
+    echo "Starting tracy-capture..."
+    tracy-capture -f -o "$OUTPUT_DIR/tracy-profile.tracy" &
+    TRACY_PID=$!
+    sleep 0.5  # give tracy-capture time to connect
+  fi
+
+  BB_BENCH_ARGS=(--reth-new-payload --wait-for-persistence)
+  if [ -n "${BENCH_WAIT_TIME:-}" ]; then
+    BB_BENCH_ARGS+=(--wait-time "$BENCH_WAIT_TIME")
+  fi
+  # Limit number of payloads if blocks count is specified
+  if [ "${BENCH_BLOCKS:-0}" -gt 0 ] 2>/dev/null; then
+    BB_BENCH_ARGS+=(--count "$BENCH_BLOCKS")
+  fi
+
+  echo "Running big blocks benchmark (replay-payloads)..."
+  $BENCH_NICE "$RETH_BENCH" replay-payloads \
+    "${BB_BENCH_ARGS[@]}" \
+    --payload-dir "$BIG_BLOCKS_DIR/payloads" \
+    --engine-rpc-url http://127.0.0.1:8551 \
+    --jwt-secret "$DATADIR/jwt.hex" \
+    --output "$OUTPUT_DIR" 2>&1 | sed -u "s/^/[bench] /"
+else
+  # Standard mode: warmup + new-payload-fcu
+  # Warmup
+  $BENCH_NICE "$RETH_BENCH" new-payload-fcu \
+    --rpc-url "$BENCH_RPC_URL" \
+    --engine-rpc-url http://127.0.0.1:8551 \
+    --jwt-secret "$DATADIR/jwt.hex" \
+    --advance "${BENCH_WARMUP_BLOCKS:-50}" \
+    "${EXTRA_BENCH_ARGS[@]}" 2>&1 | sed -u "s/^/[bench] /"
+
+  # Start tracy-capture after warmup so profile only covers the benchmark
+  if [ "${BENCH_TRACY:-off}" != "off" ]; then
+    echo "Starting tracy-capture..."
+    tracy-capture -f -o "$OUTPUT_DIR/tracy-profile.tracy" &
+    TRACY_PID=$!
+    sleep 0.5  # give tracy-capture time to connect
+  fi
+
+  # Benchmark
+  $BENCH_NICE "$RETH_BENCH" new-payload-fcu \
+    --rpc-url "$BENCH_RPC_URL" \
+    --engine-rpc-url http://127.0.0.1:8551 \
+    --jwt-secret "$DATADIR/jwt.hex" \
+    --advance "$BENCH_BLOCKS" \
+    "${EXTRA_BENCH_ARGS[@]}" \
+    --output "$OUTPUT_DIR" 2>&1 | sed -u "s/^/[bench] /"
+fi
+
+# cleanup runs via trap