Summary

Discord WebSocket connection drops approximately every 10 minutes on a stable wired 1Gbps fiber connection. Messages sent during the ~30 second reconnect window are silently lost — no retry, no delivery after reconnect.

Environment

• OpenClaw version: 2026.3.13
• OS: macOS (Darwin 25.3.0, arm64)
• Gateway mode: local (loopback)
• Network: Wired 1Gbps fiber (not Wi-Fi)
• VPN: Tailscale (recommended by OpenClaw setup wizard)

Log Evidence

From /tmp/openclaw/openclaw-2026-03-20.log, the health monitor fires repeatedly:

[gateway/health-monitor] [discord:default] health-monitor: restarting (reason: disconnected)

Observed at 08:27, 09:10, 09:20, 09:22 MDT — roughly every 10 minutes. Each restart triggers the full Discord provider init sequence (fetch-application-id, deploy-commands, WebSocket reconnect), taking ~2 seconds total.

Impact

Messages sent during the reconnect window are permanently lost. The bot does not replay missed events after reconnecting. This causes real reliability issues — user sends a message, gets no response, assumes the bot is offline. Happened multiple times today across multiple channels (#general, #restingsycamore).

Expected Behavior

One or more of:

1. Faster reconnect (sub-second) using Discord's RESUME flow instead of full re-init
2. Replay missed messages/events after reconnect using Discord's seq tracking
3. At minimum, a catch-up read of recent channel messages on reconnect so missed messages aren't silently dropped

Notes

• This is on a rock-solid wired connection — not a network quality issue
• The disconnect pattern is very regular (~10 min intervals), suggesting it may be Discord's server-side idle timeout rather than a client-side issue
• Tailscale VPN is running but was recommended by the OpenClaw setup wizard