The call to wait for no oplog holes on single-node replica sets can take global locks. This overla with the write concern timer that is active at that point in the code.

To properly understand where we are spending time, we should fix this problem or include waitForWriteConcernMillis in the workingMillis calculation