At stepdown or shutdown, ReplicationCoordinatorExternalStateImpl::_stopAsyncUpdatesOfAndClearOplogTruncateAfterPoint calls _replicationProcess->getConsistencyMarkers()->setOplogTruncateAfterPoint(opCtx, Timestamp()) and the comment about there not being any oplog holes to worry about locking for is correct, but setOplogTruncateAfterPoint() also resets _lastNoHolesOplogTimestamp and _lastNoHolesOplogOpTimeAndWallTime. Stepdown and shutdown could concurrently call _stopAsyncUpdatesOfAndClearOplogTruncateAfterPoint() resulting in a double destruction.