Learn how Replication Server resolves deadlocks.
If a transaction is ready to commit, but cannot because it is not next in proper commit order, and this transaction is holding locks on resources that are needed by a transaction that must commit before this one, a database resource deadlock occurs at the replicate database.
The database resource deadlock consists of the lock on rs_threads held by the next transaction in commit order, and the locks held on resources needed by that transaction. The database resource deadlock is detected by the replicate database, which chooses a transaction to roll back.
Since Replication Server must guarantee commit order, when this rollback is forced by the replicate database, Replication Server rolls back all transactions executing against the replicate database and reapplies them serially in commit order.