Delays and failures in scheduled Grok Bot routines

Update, Oct 13: Staff said the Oct 8 issue was resolved on Oct 9. Routines were held back again for a few hours on Tuesday morning, Oct 13, US time (what happened and what to re-run).

Update, Oct 9: At 10:22 PM ET on Oct 8, staff explained that runs that look early are the previous slot's run starting late, that a backup routine waits in the same queue, and that manual runs aren't held back. One user said at 5:17 AM ET that his scheduled runs worked this morning. Staff haven't posted an all-clear.

• Status: Resolved. On Oct 13, staff said the Oct 8 issue was resolved on Oct 9 (174125/31).
• First seen: Oct 8, about 10 AM ET, according to staff. The first forum report was posted at 3:55 PM ET.
• Symptoms: Scheduled routines start late, skip runs, or show Failed in Run history. Messages to your Bots still go through right away.
• Workaround: Message the Bot and ask it to run the task now, or start the routine manually. Staff said "manual runs aren't held back the way scheduled ones are."

Since about 10 AM ET on Oct 8, scheduled Grok Bot routines have been held back, running late, or getting marked as failed, staff said on the Cursor forum. While one run waits, the next slots are skipped rather than stacked up, so a 5-minute routine can go 1.5 to 2 hours without a run.

At 4:12 PM ET, staff member deanrie said routines were running again, then corrected that eight minutes later: runs were being held back again "while we deal with very high demand on our side."

Why it matters#

Jobs you scheduled for today, such as inbox triage or market-hours checks, may not have run, and staff said missed runs won't be replayed automatically. Troubleshooting won't help, and two users said it cost them usage.

How to use it#

1. Leave your routines switched on. Staff said pausing, resuming, editing the schedule, or recreating them won't help.
2. For anything time-sensitive, message the Bot directly and ask it to run that task now. An example prompt:

Run my "Inbox triage" routine now, and send me a one-line result when it's done.

3. Check Run history for the slots that were missed today, and rerun the ones that matter the same way.
4. Don't count on a backup routine. Staff said a backup "waits in the same queue as the primary, so it can't step in when the primary is held."
5. Watch forum topic 174125. Staff said they'll post there once routines run normally again.

Good to know#

• A run that fires 20 minutes early is an earlier slot's run going out late, staff said. The 12:07 run in one report "was your 10:31 run finally going out."
• Don't use the status page as the signal. Staff said a resolved notice there on Oct 8 "was referring to a different issue."
• Routines have queued before. In September, US morning runs started 20 to 45 minutes late until staff added capacity on Sep 24.
• Staff have suggested asking each routine to always send a one-line result, so a quiet run doesn't look like a missed one.
• We didn't reproduce this server-side problem. This entry relies on staff posts.