The CSRS catalog manager has write retry logic in order to handle transient replication error conditions. We currently do not retry WriteConcernFailed errors and will propagate them back to the caller.

This is visible in some of the continuous stepdown suite failures.