org.jboss.mq.pm.jdbc2.PersistenceManager does contain a simple logic in method remove(MessageReference, Tx) to retry a DB operation and therefore bypass sporadic temporary DB failures.
This should be extended to the other DB-operations of PersistenceManager.
Tests using a JDBC-driver throwing randomly arbitrary SQLExceptions showed this resolving most of them and keeping the system in a consistent state.
Tests without a patched PersistenceManager ran quite fast into a dead end where JMS messages were not being processed anymore until restart of the AS - also see http://www.jboss.org/index.html?module=bb&op=viewtopic&t=151023 for this symptom.
Proposed patch will get attached.