Bug type

Regression (worked before, now fails)

Summary

In OpenClaw 2026.3.8, cron jobs are accepted, scheduled, and their timers fire, but isolated agentTurn cron jobs do not complete. Even trivial one-shot isolated jobs hang until the cron timeout and then fail with cron: job execution timed out.

Steps to reproduce

I tested two one-shot isolated cron jobs.

Repro 1: minimal isolated reply job

{
"name": "cron-diagnostic-isolated-minimal",
"schedule": {
"kind": "at",
"at": "2026-03-09T22:01:00+08:00"
},
"payload": {
"kind": "agentTurn",
"message": "Reply with exactly: CRON_ISOLATED_OK",
"thinking": "low",
"timeoutSeconds": 120
},
"sessionTarget": "isolated",
"enabled": true,
"delivery": {
"mode": "announce",
"channel": "last",
"bestEffort": false
}
}

###Repro 2: isolated file-write job

{
"name": "cron-diagnostic-isolated-filewrite",
"schedule": {
"kind": "at",
"at": "2026-03-09T22:02:00+08:00"
},
"payload": {
"kind": "agentTurn",
"message": "Write a file at /tmp/openclaw-cron-diagnostic.txt containing exactly one line: CRON_FILE_OK_2026-03-09 . Then reply with exactly: CRON_FILE_DONE",
"thinking": "low",
"timeoutSeconds": 120
},
"sessionTarget": "isolated",
"enabled": true,
"delivery": {
"mode": "announce",
"channel": "last",
"bestEffort": false
}
}

Expected behavior

the minimal reply job should complete quickly and announce 'CRON_ISOLATED_OK'
the file-write job should create /tmp/openclaw-cron-diagnostic.txt and announce 'CRON_FILE_DONE'

Actual behavior

both jobs fail with:
'cron: job execution timed out'
'/tmp/openclaw-cron-diagnostic.txt' is not created
delivery status remains unknown

OpenClaw version

2026/3/8

Operating system

Ubuntu 24.04 LTS

Install method

npm global

Logs, screenshots, and evidence

cron runs results
Minimal isolated reply job:

{
"entries": [
{
"ts": 1773064980030,
"jobId": "77995402-c274-475d-99c3-1ec1adbea393",
"action": "finished",
"status": "error",
"error": "cron: job execution timed out",
"runAtMs": 1773064860009,
"durationMs": 120003,
"deliveryStatus": "unknown"
}
]
}

Isolated file-write job:

{
"entries": [
{
"ts": 1773065102057,
"jobId": "5e2b4b1a-565e-4ecd-a62c-5f4c0ec1b578",
"action": "finished",
"status": "error",
"error": "cron: job execution timed out",
"runAtMs": 1773064982045,
"durationMs": 120004,
"deliveryStatus": "unknown"
}
]
}

Impact and severity

This does not look like a scheduler-timing bug anymore.

The scheduler appears to:

register jobs
arm timers
trigger runs on time
persist failed run records
The broken layer appears to be execution of isolated cron jobs with:

{
"sessionTarget": "isolated",
"payload": {
"kind": "agentTurn"
}
}

Those runs appear to hang until timeout instead of completing.

Additional information

No response