[BUG] Cowork scheduler stops firing recurring tasks after ~24-30 hrs uptime; only Windows reboot recovers

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

The Claude desktop app's scheduled-task scheduler stops firing recurring (cron-based) tasks after some indeterminate uptime period. One-time fireAt tasks scheduled within the stall window also fail to fire. Restarting the Claude app does NOT recover. Only a full Windows OS reboot resets the scheduler.

What Should Happen?

Recurring scheduled tasks should fire reliably on their cron schedule for the lifetime of the user's session. The scheduler service should:

(a) self-heal if it stalls — a watchdog that detects missed firings and recovers without requiring a Windows reboot
(b) surface errors to the user when stalls do happen, rather than failing silently
(c) recover via app restart at minimum — full OS reboot should not be required
(d) provide diagnostic logging users can access (logs from the scheduler service, missed-fire events queued for catch-up)

A user with 60+ scheduled tasks should be able to depend on them firing within a few minutes of their scheduled time, every time, for as long as the app is running.

Error Messages/Logs

Occurrence 1 (2026-04-29 evening → 2026-04-30 night, ~24-30 hrs)**
- Last successful fire: hourly-email-triage at 4/29 13:02 UTC
- App restart attempted: did NOT resolve
- Full Windows restart: resolved. Catch-up batch fired at 5/1 04:07 UTC.

**Occurrence 2 (2026-05-01 daytime → 2026-05-02 morning, ~24+ hrs)**
- Last successful fire: daily-brief at 5/1 12:06 UTC
- Discovered: 5/2 morning when daily-brief failed to fire at 7:06 AM CT
- Possibly correlated with: PowerShell cleanup script run Fri afternoon (cleared temp folders, restarted Windows Explorer)
- Status: Windows reboot pending

## Diagnostic notes

- `/doctor` slash-command returns "not available in this environment" in Cowork — diagnostic tooling gap between Code CLI and Cowork desktop
- No visible error to user; scheduler just goes silent

Steps to Reproduce

Tasks affected: ALL recurring (e.g., daily-brief, hourly-email-triage, eod-triage, daily-crypto-brief, weekly-market-context-sunday, crypto-price-alerts)

• One-time tasks scheduled within the stall window also fail to fire
• After Windows reboot: recurring tasks catch-up-fire (multiple at once) and resume normal cadence
• Restarting just the Cowork/Claude app is NOT sufficient — must be full OS reboot

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

Claude 1.5354.0 (9a9e3d) 2026-04-29T01:14:34.000Z

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

_No response_