Summary

On macOS, openclaw-gateway enters a restart loop because notifyActiveTaskWaiters() throws:

TypeError: undefined is not iterable (cannot read property Symbol(Symbol.iterator))

This causes the gateway to disconnect repeatedly.

Version

• OpenClaw: 2026.4.5
• Runtime: Node 25.6.1
• OS: macOS

Symptom

The gateway appears to come up normally, then crashes and restarts every ~20-40 seconds.

Observed log pattern:

[diagnostic] lane task error: lane=main durationMs=...
[openclaw] Unhandled promise rejection: TypeError: undefined is not iterable (cannot read property Symbol(Symbol.iterator))
    at Array.from (<anonymous>)
    at notifyActiveTaskWaiters (file:///.../command-queue-Cssp02gj.js:88:29)
    at file:///.../command-queue-Cssp02gj.js:136:8
[gateway] loading configuration…
[gateway] ready (...)

Stack Trace

TypeError: undefined is not iterable (cannot read property Symbol(Symbol.iterator))
    at Array.from (<anonymous>)
    at notifyActiveTaskWaiters (file:///Users/mini/.npm-global/lib/node_modules/openclaw/dist/command-queue-Cssp02gj.js:88:29)
    at file:///Users/mini/.npm-global/lib/node_modules/openclaw/dist/command-queue-Cssp02gj.js:136:8

What I already checked

• This is not only an auth/token mismatch issue.
• I initially saw older openclaw-tui clients reconnecting with token mismatch, but after clearing those, the restart loop still continued.
• I marked stale running sessions as done in:
  • ~/.openclaw/agents/main/sessions/sessions.json
• After that, there were no remaining obvious stuck running sessions, but the gateway still restarted.

Likely Root Cause

notifyActiveTaskWaiters() assumes queueState.activeTaskWaiters is iterable:

for (const waiter of Array.from(queueState.activeTaskWaiters)) { ... }

In my case, queueState.activeTaskWaiters appears to become invalid/corrupted at runtime, so Array.from(...) itself throws before waiter validation can even run.

This looks like a command-queue / restart / task-drain lifecycle bug rather than a channel-specific problem.

Temporary Local Mitigation

I added a local guard in command-queue-Cssp02gj.js to rebuild activeTaskWaiters if it is no longer a Set, and to drop malformed waiter entries.

Conceptually:

const activeTaskWaiters =
  queueState.activeTaskWaiters instanceof Set
    ? queueState.activeTaskWaiters
    : (queueState.activeTaskWaiters = new Set());

After that patch, the gateway stopped immediately crashing and remained up long enough to process inbound WeCom traffic normally.

Why I think this is related to restart/task lifecycle

I noticed recent release notes mention fixes in nearby areas:

• gateway reload / restart loop behavior
• task registry maintenance affecting gateway event loop
• interrupted session wake after restart

So this may be another bug in the same lifecycle surface.

Environment Details

• launchd service: ai.openclaw.gateway
• install path:
  • /Users/mini/.npm-global/lib/node_modules/openclaw/dist/command-queue-Cssp02gj.js

Additional Info

I can also provide:

• the relevant sessions.json state before cleanup
• the repeating gateway log window
• confirmation that a defensive guard around activeTaskWaiters stops the restart loop

Question

Is this a known issue in 2026.4.5? If not, I can help provide a more minimal repro or test a candidate patch.