PR #80160: fix: add actionable queue failure diagnostics

• Repository: openclaw/openclaw
• Author: markus-lassfolk
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/80160

Description (problem / solution / changelog)

Summary

Fixes openclaw/openclaw#80156 by making failed embedded-run message queue attempts actionable in logs/diagnostics.

Instead of the low-context debug line:

queue message failed: sessionId=<id> reason=no_active_run

this now emits a warning with:

• session id
• reason (no_active_run, not_streaming, compacting)
• active embedded run count
• known session key when available
• snapshot status/run id when available
• a hint clarifying that stale/completed runs are usually not gateway outages

Why

Operators can currently receive generic alerts like:

⚠️ 🔌 Gateway: agents failed

while the gateway itself is healthy. The useful cause is buried in journald. This patch makes the underlying queue failure diagnostic self-explanatory so notification layers and operators can distinguish stale-run follow-ups from real gateway/agent failure.

Tests

Added a regression test for queueEmbeddedPiMessage when queueing into a missing run, asserting the log includes session id, reason, active count, and non-outage hint.

Local validation

I could not run the full vitest target in the sparse checkout because dependencies were not installed and /tmp is space constrained. I did run:

git diff --check

and this PR should be validated by upstream CI.

Changed files

• src/agents/pi-embedded-runner/runs.test.ts (modified, +16/-0)
• src/agents/pi-embedded-runner/runs.ts (modified, +19/-3)