The overall problems is that JournalHashMap is using a local journal most of the time and never the replicated journal.

In this line:

https://github.com/apache/activemq-artemis/blob/38693370c962f9a4d7fbc9dbeb7bffba18f9ad41/artemis-protocols/artemis-amqp-protocol/src/main/java/org/apache/activemq/artemis/protocol/amqp/connect/mirror/AckManager.java#L82

We send the current journal.. and if a switch happens inside the JournalStorageManager the switch is never captured, and we will never replicate pending acks.

When a failover happens the replica will not have the retries and the system will not reattempt them.