LineWrapper.splitTokens handles the cases I can think of

Basically... splitTokens will never put whitespace at the beginning of
wrapped display line. It also splits tokens without regard for
whitespace if the token has no whitespace but is too wide to display on
a single line.

src/atom/line-wrapper.coffee

@@ -42,22 +42,46 @@ class LineWrapper
     while tokens.length
       nextToken = tokens[0]
       if length + nextToken.value.length > @maxLength
-        # keep any leading whitespace on current line
-        if match = /\b/.exec(nextToken.value)
-          if match.index > 0
-            tokens[0..0] = @splitTokenAt(nextToken, match.index)
-          else
-            break
+        tokenFragments = @splitBoundaryToken(nextToken, @maxLength - length)
+        [token1, token2] = tokenFragments
+        tokens[0..0] = _.compact(tokenFragments)
+        break unless token1
       nextToken = tokens.shift()
       length += nextToken.value.length
       screenLine.push nextToken
 
     [screenLine].concat @splitTokens(tokens)
 
-  splitTokenAt: (token, index) ->
-    { type, value} = token
-    value1 = value.substring(0, index)
-    value2 = value.substring(index)
+  splitBoundaryToken: (token, boundaryIndex) ->
+    { value } = token
+
+    # if no whitespace, split it all to next line if it will fit.
+    # if it's longer than the max width, chop it without regard for whitespace.
+    unless /\s/.test(value)
+      if value.length > @maxLength
+        return @splitTokenAt(token, boundaryIndex)
+      else
+        return [null, token]
+
+    # if only whitespace, keep it all on current line.
+    return [token, null] unless /\w/.test(value)
+
+    # if words + whitespace, try to split on start of word closest to the boundary
+    wordStart = /\b\w/g
+
+    while match = wordStart.exec(value)
+      breakIndex = match.index
+      break if breakIndex > boundaryIndex
+
+    # if the only word start is at the beginning of the token, put the whole token on the next line
+    return [null, token] if breakIndex == 0
+
+    @splitTokenAt(token, breakIndex)
+
+  splitTokenAt: (token, splitIndex) ->
+    { type, value } = token
+    value1 = value.substring(0, splitIndex)
+    value2 = value.substring(splitIndex)
     [{value: value1, type }, {value: value2, type}]
 
   buildWrappedLineForBufferRow: (bufferRow) ->