1) Deploy a datasource not using the java JNDI context 2) Run a client which looks up the remote datasource and then spawns multiple threads which repeatedly run trivial queries. Shortly after you will likely state getting errors about connections or statements not being found, and potentially experience high cpu usage by JBoss if threads get stuck in HashMap.put()

Datasources exposed remotely via RMI used the maps of connections and statements in a way that was not threadsafe. This resulted in either clients receiving an error message "unable to find statement" or the server getting stuck in an infinite loop due to the map data structure being corrupted. This issue has been resolved by changing the method of access to the remote datasource's map, so that it is thread-safe.