In case 2 requests/threads have a db transaction that write to the same row simultaneously, one will fail. The HTTP request for which the transaction failed will be reprocessed for at most 4 times. In case this happens for a request in which a programming job has started, the execution of this job in the first or retried request may fail, and the lifecycle of the job becomes unpredictable.
Fix: skip starting the the same job during a retry of the request, in case the job number (session label) is already known in the backend.

Submitted by Elmer van Chastelet on 24 February 2021 at 10:07