It is possible for timestamped_reads_wait_for_prepare_oplog_visibility.js to fail if the writes we do are ahead of the all committed timestamp and do not make it into the transaction. This is much more likely in suites like logical_session_cache_replication_100ms_refresh because that suite runs a lot of background operations like the LogicalSessionCacheReaper that could hold back the write from making it into the all committed.

We should have these inserts be written with majority write concern to guarantee that they will make it into the all committed snapshot.