Particularly during secondary oplog application, if a write conflict is repeatedly hit, it indicates a problem that is not going to go away. Right now, we don't have a lot of diagnostic information to help understand what's going on when a write conflict is hit. After we see a few thousand write conflicts in a row for the same transaction, we should make use of some diagnostic APIs that WT has to dump out sessions and other global state. This could help us piece together what could be going on.