Summary

Agents occasionally stop responding to messages entirely. During these episodes, the gateway error log shows repeated EventQueue listener timed out after 30000ms for both MESSAGE_CREATE and INTERACTION_CREATE events. The typing indicator also hits its 2-minute TTL and stops. Recovery is eventually automatic (a few minutes), or a gateway restart clears it faster.

Observed Pattern

• Agent goes silent — stops responding to all incoming messages
• Switching models during the episode doesn't help
• Gateway error log: [EventQueue] Listener DiscordMessageListener timed out after 30000ms for event MESSAGE_CREATE
• Typing TTL log: typing TTL reached (2m); stopping typing indicator
• Typically resolves on its own after a few minutes, or immediately after openclaw gateway restart
• Often coincides with large session files being processed (e.g. a 20MB .reset file being handled during compaction)

Likely Cause

Session compaction/archiving appears to involve synchronous file I/O (readFileSync, renameSync) on large files, which can block the Node.js event loop long enough for the 30-second EventQueue listener timeout to fire. Once the event loop catches up, the agent recovers.

Suggested Fix

Moving session compaction and large file operations to a worker thread (or using async streaming reads) would keep the event loop free during heavy I/O, preventing the timeout cascade.

Workaround

openclaw gateway restart immediately unblocks the agent. Otherwise it recovers on its own once the blocking I/O completes.

Environment

• OpenClaw: 2026.2.26
• macOS 26.2 (arm64)