The new test added in SERVER-44711 for metrics.repl.network.oplogGetMoresProcessed requires_document_locking. This is because the test blocks oplog fetching on a failpoint while trying to do oplog writes. Under ephemeralForTest, the oplog writes (acquiring X lock) would conflict with the oplog reads (holding S lock) that's blocked on a failpoint.

Normally that's not a problem because oplog writes should take MODE_IX locks and oplog reads should take MODE_IS when the storage engine supports document level locking.