Environment

• OpenClaw: 2026.4.15
• OS: Windows 11
• Gateway mode: local, bind: lan
• Model provider: Ollama (remote via direct Ethernet)

Steps to reproduce

1. Gateway running normally
2. Edit (e.g. add )
3. Gateway detects the change and triggers hot-reload via SIGUSR1
4. Old process does not fully exit before new process tries to bind port 18789
5. New process fails, exits immediately
6. Lock file is left behind
7. All subsequent restart attempts fail — port binds but WebSocket never completes handshake (1006 abnormal closure)
8. Only recovery: manually delete lock file, then restart

Expected behavior

Config hot-reload should complete cleanly, or gracefully handle the case where SIGUSR1 is unsupported on Windows (currently logs ).

Workaround

Stop gateway → edit config → restart. Added to as a safety net.

Related

• [Bug]: SIGUSR1/config.patch restart orphans gateway under custom supervisors → EADDRINUSE crash loop (workaround: OPENCLAW_NO_RESPAWN=1) #65668 (SIGUSR1 restart orphans gateway under custom supervisors)
• [Bug]: Gateway health check fails (1006) + Ollama model timeout during onboarding on Windows #61805 (Gateway health check 1006 on Windows)