Details

Make job runners commit changes for replica DB-lag prone jobs one job at a time. This is useful if there are many job workers that race on replica DB lag checks. If set, jobs taking this many seconds of DB write time have serialized commits.

Note that affected jobs may have worse lock contention. Also, if they affect several DBs at once they may have a smaller chance of being atomic due to the possibility of connection loss while queueing up to commit. Affected jobs may also fail due to the commit lock acquisition timeout.