Issues

Issues

• WebSocket upgrades never reuse a warm isolate, and placement.region spreads requests across many isolates

  cloudflare/workerd#8379 · opened on Nov 12, 2026

• SIGTRAP at shutdown: CHECK_EQ(reference_count_, 1) in IsolateGroup::ReleaseDefault (TearDownOncePerProcess), refcount=2 — workerd 1.20261023.1

  cloudflare/workerd#8370 · opened on Nov 11, 2026

• new_module_registry increases Worker startup time; about 10 ms of it appears only with nodejs_compat

  cloudflare/workerd#8365 · opened on Nov 11, 2026

• cluster: alarm retry limit and backoff stop working after a cold wake, and a throwing constructor loops with no delay

  cloudflare/workerd#8351 · opened on Nov 11, 2026

• cluster: synchronous JS or Wasm longer than about 12 s exits the process, and one object can take down every node

  cloudflare/workerd#8350 · opened on Nov 11, 2026

• fix(types): ForwardableEmailMessage omits the documented canBeForwarded property

  cloudflare/workerd#8290 · opened on Nov 8, 2026

• D1 regression since 2026-11-08: "VIEWs and/or subqueries nested too deep" (SQLITE_ERROR 7500) when a CTE is referenced more than ~100 times in one statement

  Bug

  cloudflare/workerd#8389 · opened on Nov 8, 2026
  7

• TCP sockets: allow setting ALPN on connect() and startTls() (needed for PostgreSQL 17+ direct TLS)

  cloudflare/workerd#8287 · opened on Nov 8, 2026

• node:buffer: Buffer.alloc(size, fill, encoding) throws when fill is a number or a Buffer

  cloudflare/workerd#8286 · opened on Nov 8, 2026

• Streaming TextDecoder drops U+FEFF at the start of a chunk that follows an ASCII-only chunk

  cloudflare/workerd#8279 · opened on Nov 7, 2026
  1

• Diagnosing hung RPC invocations when a WorkerEntrypoint forwards a Durable Object ReadableStream (Cloud observation)

  cloudflare/workerd#8278 · opened on Nov 7, 2026

• 🐛 Bug Report — cluster: a large SQLite transaction fails with "capture lost"

  cloudflare/workerd#8275 · opened on Nov 7, 2026

• 🐛 Bug Report — Runtime APIs: Synchronized 1006 WebSocket disconnects followed by same-version Durable Object restart

  cloudflare/workerd#8244 · opened on Nov 5, 2026

• 🐛 Bug Report — Runtime APIs: Hosted hibernatable WebSocket close callbacks delayed up to 17 minutes after server close requests

  cloudflare/workerd#8243 · opened on Nov 5, 2026

• Promises created in optimized code have no context tag, so a request awaiting one resumes in the wrong request

  cloudflare/workerd#8228 · opened on Nov 5, 2026

• Bug Report — Runtime APIs - concurrent env.AI.run() calls can return responses in the wrong format

  cloudflare/workerd#8203 · opened on Nov 3, 2026

• Intermittent linux-x64 SIGSEGV during local D1 operations — workerd 1.20260515.1

  cloudflare/workerd#8190 · opened on Nov 3, 2026

• Cold hibernating Durable Object remains physically closing after a server-initiated Close

  cloudflare/workerd#8185 · opened on Nov 3, 2026

• Cluster mode: deploying new JS and upgrading workerd without downtime

  cloudflare/workerd#8155 · opened on Nov 2, 2026
  1

• Inspector cannot attach to WorkerLoader names containing '/'

  cloudflare/workerd#8142 · opened on Nov 2, 2026

• 🐛 Bug Report — Runtime APIs: node:crypto GCM decipher requires setAuthTag() before update(); Node allows it any time before final()

  cloudflare/workerd#8123 · opened on Nov 1, 2026

• macOS: acceptAuthenticated fails with ENOTCONN when a Unix peer disconnects before accept

  cloudflare/workerd#8122 · opened on Nov 1, 2026

• Only compiles on Debian

  cloudflare/workerd#8104 · opened on Oct 30, 2026

• Outbound fetch fails when a pooled keep-alive connection is closed by the server (flaky fetch/misc@all-autogates)

  cloudflare/workerd#8100 · opened on Oct 30, 2026
  2

• Local dev containers on macOS: another process on 127.0.0.1 can take the egress sidecar's port ("Updating sidecar egress config failed with: 501")

  cloudflare/workerd#8081 · opened on Oct 29, 2026