**OpenClaw version:** 2026.4.5 (3e72c03)
**Mac:** Darwin 24.6.0 (arm64)
**Node:** v25.8.1
**Model:** MiniMax-M2.7

Bug Description

Scheduled cron jobs consistently time out at ~60s, but running the same job manually via openclaw cron run <id> completes successfully in ~20-25s.

All 4 cron jobs are affected (sf-sync-contracts-form, sf-daily-contracts, wocentral-invoice-check, session-cleanup).

Steps to Reproduce

1. Create a cron job targeting an isolated session (no agentId specified, or agentId blank)
2. Wait for scheduled run — it times out at ~60s
3. Run openclaw cron run <id> --timeout 300000 manually — it completes successfully in ~20s

Expected Behavior

Scheduled runs should complete the same way manual cron run does.

Run History

Manual run (sf-sync-contracts-form, run just now):

• Status: ok
• Duration: 22,124ms
• Output: 8,534 contracts synced, Netlify deploy succeeded

Scheduled runs (same job):

• Status: error
• Duration: ~60,000-83,000ms
• Error: "Request timed out before a response was generated"
• Output tokens: only 118 (vs 461+ on successful manual runs)

Jobs Config

ID                                   Name                     Target    Agent ID
a3f7c892-1d5e-4b8a-9c6f-2e8d1a4b7c3f sf-sync-contracts-form   isolated  -
bfe46c46-25b0-4699-912c-db2e2c7c4f6e session-cleanup          isolated  main (succeeds when manually triggered too)
11957982-78ef-42f0-9b2d-cb7f96c469e6 sf-daily-contracts       isolated  -
04f76353-d9ad-4adf-a8f0-de0ab7087ff4 wocentral-invoice-check  isolated  -

Notes

• This started around Apr 4-5, 2026 (before that, scheduled runs were succeeding)
• agents.defaults.timeoutSeconds: 600 is set in config
• The session-cleanup job (which has agentId: main) also times out when scheduled, but succeeds when manually run
• The issue appears to be in how the cron scheduler delivers responses to isolated sessions — the response stream seems to get cut at ~60s regardless of config