This is a test-fix only ticket. The issue only happens to MMAPV1 on v3.6 and v4.0.
lock_stats_suboperation_logs.js scans every line from the logs and captures any lock waiting stats. The assumption is that only two lines will be captured: they are the logs for the sub-command and the parent command. Also, their lock waiting time should be the same.
However, the assumption does not always hold: the logs being scanned include logs from validation hooks and MMAPV1 JournalFlushLock are taken during this stage. It's possible that there are some wait for this lock so the test failed due to these noise lock stats.