Description

After upgrading from 2026.2.26 to 2026.3.1, the Discord provider enters a stuck-restart loop where the health-monitor flags it as "stuck" every ~10 minutes, despite Discord successfully logging in and resolving channels each time.

Environment

• OpenClaw version: 2026.3.1 (upgraded from 2026.2.26)
• OS: macOS 26.3 (arm64), Mac mini M4, 16GB
• Node: v25.5.0
• Discord config: 1 guild, ~50 channels, groupPolicy: allowlist, streaming: off, threadBindings enabled

Reproduction

1. Upgrade from 2026.2.26 to 2026.3.1 via npm i -g openclaw@latest
2. Gateway restarts via doctor/SIGUSR1
3. Discord logs in successfully, resolves all channels and users
4. ~10 minutes later, health-monitor flags Discord as stuck and restarts it
5. Loop repeats indefinitely

Logs

2026-03-02T11:51:54.063Z [health-monitor] [discord:default] health-monitor: restarting (reason: stuck)
2026-03-02T11:51:54.485Z [discord] [default] starting provider (@OCM Ops)
2026-03-02T11:51:55.424Z [discord] logged in to discord as 1470138329826590792 (OCM Ops)
2026-03-02T12:01:54.087Z [health-monitor] [discord:default] health-monitor: restarting (reason: stuck)
2026-03-02T12:11:54.102Z [health-monitor] [discord:default] health-monitor: restarting (reason: stuck)
2026-03-02T12:21:54.122Z [health-monitor] [discord:default] health-monitor: restarting (reason: stuck)
2026-03-02T12:31:54.143Z [health-monitor] [discord:default] health-monitor: restarting (reason: stuck)
... (19 total stuck restarts before manual intervention)

Other providers (Telegram, Slack, BlueBubbles) also occasionally flagged as stuck but less frequently.

Impact

• Messages sent to Discord channels during the stuck window are silently dropped
• Cron jobs that try to deliver messages cross-context fail with "Message failed"
• Bot appears unresponsive in Discord despite config being correct

Workaround

A full gateway restart (openclaw gateway restart or SIGUSR1) appears to break the loop temporarily. The health-monitor then stays stable. Rolling back to 2026.2.26 is the safe fallback.

Expected Behavior

Health-monitor should not flag Discord as stuck when it successfully logs in, resolves channels, and is ready to receive messages.