Due to an internal misconfiguration, between October 15, 11:00 UTC and October 16, 6:30 UTC, a small number of Actors were unable to start, either when started via API, or when scheduled to run regularly. We have identified the issue and implemented a fix, and the Actors are now starting normally.
There was no data loss, and no excessive charges for the failed Actors.
We apologize for the inconvenience, and we will contact the affected users soon.