We're seeing the exact same issue under a nearly identical configuration. Our DBA is gone for the day so I can't ask him what fix he put in place, but I can nearly guarantee that the problem is the configuration of Oracle (or the objects therein, rather), not JBoss Messaging. I suspect it has to do with fragmentation of the lobsegment. Our JBM_MSG sees about 20 million messages come and go per day, so there's lots of volatility in that table, which is a key ingredient for fragmentation.