Cron sessions not closing after agent turn completes

Summary:
Cron job sessions remain open indefinitely after the agent completes its turn, instead of auto-closing when the work is done.

Steps to reproduce:

1. Configure a cron job with agentTurn payload
2. Let it run (agent completes turn successfully in ~1-2 seconds)
3. Observe session remains open for 30+ minutes with no activity

Expected behavior:
Session should auto-close after the agent's turn completes.

Actual behavior:
Session stays open indefinitely until manually killed or health monitor detects inactivity.

Example:

{
  "schedule": { "kind": "cron", "expr": "*/5 * * * *" },
  "sessionTarget": "isolated",
  "payload": {
    "kind": "agentTurn",
    "message": "Read HEARTBEAT.md if it exists..."
  }
}

Observed pattern:

1. Gateway spawns cron session → agent turn executes
2. Agent completes turn successfully (~1.3s runtime)
3. Gateway writes final assistant message to JSONL
4. Session should close but stays open
5. Session sits idle for 30+ minutes (no new messages, no activity)
6. Eventually killed by external health monitor

Impact:

• Resource leak (sessions accumulate over time)
• Requires external cleanup to prevent session buildup
• Affects all agentTurn cron jobs

OpenClaw version:
Commit: bc475f0 (backup-before-sync-20260208)

Workaround:
Add timeoutSeconds to payload forces session closure, but this is a hard timeout rather than proper cleanup.

Logs:
Session logs show clean completion (exit/finish message), but session lifecycle doesn't close the session file or remove from active sessions.