openclaw gateway probe false-negative timeout/close on loopback while gateway is healthy.

Version / Platform

• OpenClaw: 2026.3.13 (61d171a)
• OS: Windows 11
• Gateway: local loopback ws://127.0.0.1:18789
• Auth mode: token

Summary

openclaw gateway probe (and probe --verbose) consistently reports unreachable/timeouts, but gateway is actually up and usable (health checks and device RPC work).

Repro

openclaw gateway start
openclaw gateway probe --verbose

Observed

• gateway client error: Error: WebSocket was closed before the connection was established
• Connect: failed - timeout
• Sometimes: gateway closed (1000 normal closure): no close reason

Expected

If gateway is healthy/listening and RPC works, gateway probe should report reachable (or at least distinguish auth/probe-budget issues clearly).

Control checks that succeed

openclaw gateway status
openclaw health
openclaw devices list --timeout 30000

• gateway status reports listening on 127.0.0.1:18789, RPC probe OK.
• health succeeds.
• devices list succeeds and returns paired devices/scopes.

Additional notes

• Issue persists across restart attempts and clean stop/start cycles.
• probe appears to be the outlier command vs other gateway interactions.
• There are occasional PID/lock mismatch messages during force/restart flows (already running (pid ...) when no listener initially detected).