Bug Description

After multiple gateway restarts (triggered by config hot-reload via SIGUSR1), the cron scheduler sets a nextWakeAtMs timestamp that is already in the past. It then never self-corrects — all subsequent scheduled slots are silently missed until an external cron API call (e.g. cron list) forces a recalculation.

Environment

• OS: macOS 14 (arm64)
• Node: v25.5.0
• OpenClaw: latest (installed via npm)
• Cron job: 0 7-19 * * 1-5 (hourly 7am–7pm PST, weekdays)
• Session target: isolated agentTurn

Steps to Reproduce

1. Have a cron job with schedule 0 7-19 * * 1-5 (tz: America/Los_Angeles)
2. Edit workspace files multiple times between 12:00–1:00 AM PST, triggering gateway config hot-reload (SIGUSR1 restarts)
3. Gateway restarts ~8 times between 12:14–12:48 AM
4. After final restart, cron module logs: nextWakeAtMs: 1770368400000 → 1:00 AM PST (already in the past, and outside the 7-19 schedule window)
5. Gateway stabilizes and stays running from 12:48 AM onward
6. Mac stays awake (process keeping it alive)
7. 7:00 AM, 8:00 AM, 9:00 AM slots are all silently missed — zero cron activity in logs
8. Only TypeError: fetch failed (non-fatal, unrelated) appears during 7-9 AM window
9. At 9:12 AM, a manual cron list API call is made
10. Scheduler recalculates nextWakeAtMs to 1770400800000 → 10:00 AM PST (correct next slot)

Expected Behavior

The cron scheduler should:

• Detect stale nextWakeAtMs on startup/restart and recalculate to the next valid future slot
• Have a periodic self-check timer that catches missed slots
• Not require an external API call to "unstick" itself

Relevant Log Entries

# Multiple restarts between 08:14–08:48 UTC (12:14–12:48 AM PST)
cron: started  {enabled:true, jobs:2, nextWakeAtMs:1770368400000}  # 1:00 AM PST — already past!
cron: started  {enabled:true, jobs:2, nextWakeAtMs:1770368400000}  # same stale value
# ... repeated 8 times

# 15:00–17:00 UTC (7–9 AM PST): ZERO cron activity
# Only non-fatal fetch errors appear

# 17:12 UTC (9:12 AM PST): External cron API call triggers recalculation
# nextWakeAtMs jumps to 1770400800000 (10:00 AM PST) ✅

Impact

Cron jobs silently fail to fire for hours after gateway restarts. Users have no indication anything is wrong until they manually check.