Bug type

Behavior bug (incorrect output/state without crash)

Summary

When an "LLM Request timed out" error occurs while using Model A, switching to Model B does not update the model used by the heartbeat task. The heartbeat continues to use Model A instead of the newly selected Model B.

Steps to reproduce

1. Set the model to "gpt 5.2": openclaw models set openai/gpt-5.2.
2. Perform tasks using gpt-5.2 until an "LLM Request timed out" error occurs.
3. Switch the model to "minimax-portal/MiniMax-M2.5": openclaw models set minimax-portal/MiniMax-M2.5.
4. Continue working; the system correctly uses MiniMax and functions normally.
5. Wait for the automatic Heartbeat to trigger. Check the logs.

Expected behavior

The Heartbeat task should use "MiniMax-M2.5" to complete its work.

Actual behavior

The Heartbeat task still uses "gpt-5.2," resulting in another "LLM Request timed out" error.

OpenClaw version

2026.2.26

Operating system

macOS 26.3

Install method

npm global

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response