There's a very infrequent race occurring in oplog_rollover.js that is hard to diagnose without additional logging. I believe the issue is related to failing to create a new oplog stone with try-locks causing the test to hang as oplog truncation will not run as expected.