Deadlock during 'EntirePool' flush on MySQL connection error in EAP 6

Issue

A datasource pool is configured with <flush-strategy>EntirePool</flush-strategy>

1. time=1: On thread-one, an in-use connection (JDBCMysqlXAConnection - which is locked within MySQL code) detects a failure and notifies its event listeners
2. time=2: On thread-two, during connection validation, an invalid connection is detected and the thread synchronously locks and begins flushing the pool
3. time=3: On thread-one, an XAManagedConnection instance, on connection error notification, synchronously attempts to flush the pool and blocks on the pool lock
4. time=4: On thread-two, the in-progress pool flush attempts to close the JDBCMysqlXAConnection owned by thread-one and blocks on the connection lock