Today I've deployed a new version of my retry service. I think I've worked out most of the kinks, so I'm going to leave it running overnight and see how it goes. In the worst case, it might crash and die, in which case spurious failures will not get retried until I come back and fix it tomorrow morning, Europe time. In the best case, the service will just keep doing what the old one did, with two improvements: 1. If the same job keeps failing, it will give up retries. 2. If sending a retry command to GitLab itself fails, those will also be retried a few times. -Bryan "Who retries the retryer?" - Juvenal, probably