We had an issue recently where a (transactional) replicated table was replicating data as expected.
Then about 30 or so rows in the source table were not in the destination table, but other rows created after those 30 rows were replicated.
We have pretty much confirmed that users did not delete those rows.
Unfortunately we had to resolve the issue quickly and so blew away & recreated the subscription so a lot of evidence is probably gone from the crime scene.
We cant figure out what could cause 30 rows not to be replicated, yet leave replication operational.
Anyone got any ideas?