Bug

The Discord bot's WebSocket connection can enter an unrecoverable resume loop. Observed on Feb 9, 2026 starting ~9:53 AM PST.

Behavior

1. Connection closes with codes 1005/1006
2. Client attempts resume with escalating backoff (1s → 2s → 4s → 8s → 16s)
3. Resume immediately gets closed again
4. Backoff resets to 1s and cycles indefinitely — hundreds of times over 12+ hours
5. Bot appears online but is deaf to all messages

Errors observed during the loop

• connection stalled: no HELLO received within 30000ms
• AggregateError
• DNS failure: getaddrinfo ENOTFOUND gateway-us-east1-d.discord.gg

What didn't fix it

• openclaw gateway restart (SIGUSR1) — temporarily fixed it but loop resumed within minutes
• Two automated restart attempts had the same result

What fixed it

• openclaw gateway --force (full process kill + fresh start) — resolved it permanently

Likely root cause

A transient DNS/network issue caused the initial disconnect. The session became stale but the client kept trying to resume it forever instead of falling back to a fresh connection.

Suggested fix

Add a circuit breaker: after N consecutive failed resumes, abandon the session and do a full reconnect with a new identify handshake.