Desktop scheduler: per_task_limit skips retried every 60s for days (no backoff); queued run fires hours later after unpause-window

Environment

• Claude Code desktop app on Windows 11 Pro
• Recurring local scheduled task (cron 45 9,13,17 *) whose runs sometimes take 1-2+ hours

Bug

The scheduled-tasks registry (claude-code-sessions/.../scheduled-tasks.json) accumulated ~250 recordedSkips entries with reason "per_task_limit" for one task, spaced almost exactly 60 seconds apart, in continuous blocks spanning entire working days across 5-12 Aug (laptop awake). It appears that when a fire is blocked (long-running prior session / missed windows after the laptop was offline during holidays), the scheduler retries every minute indefinitely instead of backing off or dropping the missed fire.

Consequences observed

1. Days-long retry storms recorded in the registry (file grows unboundedly).
2. A backlogged fire eventually executed hours after its scheduled window, overlapping the period when the user believed the task had just been paused - i.e. a run started when nobody expected one.

Expected

• Exponential backoff or a single catch-up attempt per missed window
• A paused task should also drop any queued/backlogged fires
• recordedSkips should be capped or rotated