Bug type

Regression (worked before, now fails)

Summary

Manual cron job triggers (both via the cron run tool/API and the dashboard "Run" button) return enqueued: true / ok: true but no isolated session is ever spawned. The job silently never executes.

Scheduled (automatic) runs work fine — multiple jobs ran successfully today on their cron schedules. Only manual/forced triggers are affected.

Steps to reproduce

1. Have a working cron job that runs successfully on schedule (e.g., daily-backup with sessionTarget: "isolated", payload.kind: "agentTurn")
2. Trigger it manually via the cron tool: cron run --jobId --runMode force
3. Response: { ok: true, enqueued: true, runId: "manual:..." }
4. Check sessions_list — no cron session appears
5. Check cron runs --jobId — no new run entry
6. Check gateway logs — no mention of the manual trigger, no spawn attempt, no errors

Expected behavior

Manual trigger should spawn an isolated session and execute the job, same as a scheduled run.

Actual behavior

API returns success but the job is silently dropped. No session spawned, no run history recorded, no log output.

OpenClaw version

2026.3.8

Operating system

macOS 26.3 (arm64, Apple Silicon)

Install method

npm (pnpm)

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

• Tested with multiple jobs (daily-backup, daily-openclaw-maintenance) — same behavior on all
• Gateway restart does not fix the issue
• The dashboard "Run" button in the cron jobs section also appears non-functional (same underlying issue)
• Scheduled runs on the same jobs work correctly (e.g., daily-backup ran at 11:30 AM today in 24s, Daily Science News ran at 4 AM, etc.)
• Cron logs only show cron: timer armed every 60s — no acknowledgment of manual enqueue