This contains just some other fixes now. It seems like the job queue did actually uniquely assign jobs.
We first query for a job that is not running, and then set it to running. This would mean that if two simultaneous transactions would set the same job to running, then one of them would encounter a serialisation failure.
This contains just some other fixes now. It seems like the job queue did actually uniquely assign jobs.
We first query for a job that is not running, and then set it to running. This would mean that if two simultaneous transactions would set the same job to running, then one of them would encounter a serialisation failure.