The stop_hook_active escape makes sense, and the failure still reaches the user, but in a long session that line is easy to scroll past. I'd count how often "checks still failing after one retry" fires across sessions. If it climbs, the agent keeps hitting errors it can't fix, and I want that flagged somewhere I'll see it.