A lot of the edge cases in the oplog truncation code are exercised on failover (do we truncate successfully after lots of step-ups?). We should add some sort of check after chaos testing (failover events) to check that truncation successfully occurs and converges by some amount TBD.