Contributions: Sep 20–27, 2026

33 substantial PRs (29 merged, 7 open/draft). 4 in node, 29 in workerd.

workerdMerged#7515

ts streams identity write settlement

Fix IdentityTransformStream behavior to correctly settle write promises

workerdMerged#7510

Release the transformer once a TransformStream finishes

Release the transformer once a TransformStream finishes to prevent memory leaks

nodeMerged#66138c++semver-minorlib / src

src: add --process-timeout=N

Add --process-timeout=N flag to set a time limit on a Node.js process

workerdMerged#7509

Settle a TransformStream cancel from the writable's state

Fix TransformStream cancel to reject with the writable's store as per the spec

nodeMerged#66117c++lib / srcauthor ready

src: ensure Socket(fd) cannot bypass allow-net permission

Ensure Socket(fd) respects --allow-net permission

workerdMerged#7507

Validate a writable chunk's size only when it is enqueued

Validate writable chunk size only when enqueued to prevent incorrect error handling

workerdMerged#7504

Lock a closed or errored native stream on tee()

Lock a closed or errored native stream on tee() to prevent multiple teeing

workerdMerged#7502

Close a native stream at the handoff of its data

Close a native stream at the handoff of its data to prevent it from staying readable forever

workerdMerged#7498

Release the pipe source when validation fails after locking

Releases pipe source when validation fails after locking to prevent source from being locked indefinitely

workerdMerged#7500

Observe the pipe's AbortSignal through an abort algorithm

Registers pipe's AbortSignal through an abort algorithm to fix abort handling

workerdMerged#7501

Pin that a pipe waits for an erroring destination's in-flight write

Fixes pipe waiting for erroring destination's in-flight write to settle before backward error propagation

workerdMerged#7479

Queue the chunk before filling reads with released bytes

Queues chunk before filling reads with released bytes to follow spec

workerdMerged#7478

Keep a byobRequest held across tee() working

Keeps byobRequest held across tee() working by leaving the request valid after a fork

workerdMerged#7477

Take queued bytes whole on auto-allocate default reads

Improved default read behavior when auto-allocate chunk size is set, reducing unnecessary buffer allocations and closures.

workerdMerged#7475

Pin async iterator interleavings that follow WebIDL

Updated async iterator behavior to follow WebIDL spec, ensuring correct promise handling and next() settlements.

workerdMerged#7474

Transfer byte-stream buffers to fixed length

Fixed buffer transfer issue by using fixed length buffers, preventing internal RangeErrors and data loss.

workerdMerged#7473

Error a tee branch or detached byte stream on a fractional fill at close

Added error handling for tee branches and detached byte streams on fractional fill at close, preventing pending reads and data loss.

workerdMerged#7471

Allow tee() of a native body with an abandoned read in flight

Allow tee() of a native body with an abandoned read in flight by running the generic KJ tee operation

workerdMerged#7469

Deliver whole elements from native BYOB reads with multi-byte views

Deliver whole elements from native BYOB reads with multi-byte views by exposing elementSize in the request

workerdMerged#7468

Return an empty view from read(view) on a cancelled byte stream

Return an empty view from read(view) on a cancelled byte stream to match spec and other implementations

workerdMerged#7467

Pin where closed settles relative to the below-min tail read

Pin where closed settles relative to the below-min tail read to ensure consistent behavior with C++ and spec

workerdMerged#7465

Settle the head read first when respond() leaves a remainder

Settle the head read first when respond() leaves a remainder to maintain correct read order

workerdMerged#7426

Walk the queued stream's cursors by index and skip the pinned cursor

Optimize StreamQueue cursor iteration by walking by index and skipping the pinned cursor

workerdMerged#7298

Implement proper js rpc stream cancel propagation

Implement proper JS RPC stream cancel propagation to fix defects in stream handling

workerdMerged#7425

Buffer the compression stage's output in blocks released as they are …

Buffer compression stage output in blocks to reduce memory usage

workerdMerged#7424

Copy body bytes out as they arrive and cancel the source when consump…

Copy body bytes as they arrive and cancel the source when consumption is complete to reduce memory usage

workerdOpen#7536

Follow the spec's iterator protocol in TS ReadableStream.from

Updated ReadableStream.from() to follow the spec's iterator protocol in TS

workerdOpen#7534

Align the TS ReadableStream async iterator with WebIDL

Aligned the TS ReadableStream async iterator with WebIDL

workerdOpen#7530

Pin SharedArrayBuffer view rejection across byte-stream entry points

Pinned SharedArrayBuffer view rejection across byte-stream entry points

workerdOpen#7529

Tune re-entrancy and microtask resolution timing

Tuned re-entrancy and microtask resolution timing to align with spec and prevent user code from running in internal paths

nodeOpen#66298bufferc++semver-minor

buffer: add isLatin1

Add a fast check to determine if a string is Latin1, improving performance over regex checks

nodeOpen#66115c++semver-minorlib / src

perf_hooks: add options to monitorEventLoopDelay()

Added options to tune histogram parameters to monitorEventLoopDelay() and fixed a truncation and doc bug.

workerdDraft#7528

Add perfetto for rust

Add perfetto for Rust to improve performance monitoring