Bug type

User-facing error / background task classification

Summary

On OpenClaw 2026.5.12 (f066dd2), a background heartbeat run on main failed with a Codex app-server / command-lane timeout and was delivered to Discord with the generic foreground user-request failure text:

Something went wrong while processing your request. Please try again, or use /new to start a fresh session.

The failing lane was a heartbeat lane, not a user-initiated Discord request:

session:agent:main:heartbeat:heartbeat

From the user perspective this looks like a foreground chat request failed, even when user chat may still be available.

Environment

• OpenClaw: 2026.5.12 (f066dd2)
• Install: Homebrew-backed global package
• OS: macOS / LaunchAgent
• Surface: Discord channel
• Main model: openai/gpt-5.5 through OpenAI Codex app-server
• Heartbeat was configured with a Discord target and timeoutSeconds: 90

Private IDs, paths, and auth details are intentionally omitted/redacted.

Observed behavior

A background heartbeat run used the main agent/model and timed out. Logs showed the heartbeat lane and embedded failure shape:

lane=session:agent:main:heartbeat:heartbeat
codex app-server client retired after timed-out turn
Profile openai-codex:... timed out. Trying next account...
CommandLaneTaskTimeoutError: Command lane "main" task timed out after 120000ms
Embedded agent failed before reply: Command lane "main" task timed out after 120000ms

Discord received the generic user-request failure copy suggesting /new.

Expected behavior

Heartbeat/background failures should be classified and mapped separately from foreground user-request failures.

Reasonable behavior would be one of:

1. suppress visible heartbeat failure by default and log it as heartbeat diagnostic state, or
2. post heartbeat-specific copy only when configured, e.g.

Heartbeat check failed due to a model/runtime timeout. User chat remains available.

The generic /new guidance should not be posted for background heartbeat failures unless the affected foreground session is actually unusable.

Related issues / PRs

Related but not exact:

• Codex app-server turns time out after 60s despite meaningful progress #78756: Codex app-server turns time out after progress and surface generic channel error.
• [Bug]: Discord agent session remains routable after timeout, causing partial-success plus generic failure #72810: Discord session timeout can cause generic failure after partial success.
• feat: Add silentErrorReplies / suppressAgentErrors support for WhatsApp channel #59782 / feat: Add errorPolicy support to WhatsApp channel #59801: generic agent error suppression/customization, but currently WhatsApp-specific.
• Feature request: Persistent task-status surface for long-running channel turns #52640: persistent task-status surface for long-running channel turns.

Local mitigation

The operator mitigated locally by removing heartbeat: {} from main, keeping heartbeat on a lighter cron-worker, and keeping heartbeat isolated/light-context/skip-when-busy. That fixes this instance, but the product behavior remains confusing when any background heartbeat failure is delivered as a foreground user-request failure.