Description

Cron jobs configured with sessionTarget: main, wakeMode: next-heartbeat, and payload.kind: systemEvent stop firing after the gateway has been running for an extended period (overnight).

Steps to Reproduce

1. Configure multiple cron jobs with:

   • sessionTarget: "main"
   • wakeMode: "next-heartbeat"
   • payload: { kind: "systemEvent", text: "..." }
   • Various schedules (e.g., 0 8 * * *, 30 8 * * *, etc.)
   • Timezone configured (e.g., tz: "America/Montevideo")

2. Let the gateway run overnight without restart

3. Check cron status the next morning

Expected Behavior

Cron jobs should fire at their scheduled times and inject systemEvents into the main session.

Actual Behavior

• Crons show lastRunAtMs from the previous day
• nextRunAtMs shows times that have already passed
• Manual cron run returns { ran: false, reason: "not-due" } even when nextRunAtMs is in the past
• Gateway restart fixes the issue (recalculates schedules correctly)

Environment

• OpenClaw version: 2026.2.3-1 (d84eb46)
• OS: Linux (WSL2)
• Node: v24.11.1
• Timezone: America/Montevideo (GMT-3)

Workaround

Migrated crons to sessionTarget: isolated with payload.kind: agentTurn, which appears more reliable.

Additional Context

The scheduler seems to not recalculate nextRunAtMs correctly after midnight, possibly a timezone-related issue or a bug in the cron evaluation loop.