Description

The cron scheduler initializes correctly (logs cron: started with correct state) but jobs never actually execute. The internal timer that should fire at nextWakeAtMs doesn't trigger.

Environment

• OpenClaw version: 2026.2.3-1
• OS: macOS 15.3 (Darwin 25.2.0 arm64)
• Node: 25.5.0
• Install method: pnpm

Symptoms

1. Gateway starts, cron scheduler logs:

   cron: started {"enabled":true,"jobs":23,"nextWakeAtMs":1770394680000}
   

2. When nextWakeAtMs time arrives, nothing happens. No lane enqueue: session:agent:*:cron:* entries appear.

3. The jobs.json file IS updated with new nextRunAtMs values, but jobs don't execute.

4. Agents work fine when triggered via webchat/Telegram - only cron scheduler is broken.

5. Issue persists after:

   • Gateway restart (SIGUSR1)
   • Full reinstall (pnpm install -g openclaw)
   • Fresh gateway start

Evidence

Working (Feb 5, ~6:50 UTC):

06:49:02 - cron: started (nextWakeAtMs = 10:50)
06:50:00 - lane enqueue: session:agent:forge:cron:... ✅
06:50:16 - lane task done ✅

Broken (Feb 6, after restart):

16:17:27 - cron: started (nextWakeAtMs = 16:18)
16:18:00 - [silence - no lane enqueue] ❌
16:22:00 - [still nothing at pixel's scheduled time] ❌

Workaround Attempted

• cron.wake action with mode: "now" - no effect
• cron.run with specific jobId - returns "ran": false, "reason": "not-due"

Expected Behavior

Jobs should fire at their scheduled times, creating lane enqueue: session:agent:*:cron:* entries and executing the agent sessions.

Additional Context

The scheduler appears to be in a state where it tracks time and updates state, but the setTimeout/setInterval callback that should trigger job execution is either not set up or not firing.