A concurrency bottleneck was accidentally added to the JCA connection creation while solving other problems. This has already been fixed in EAP 6.2.0, and I'm creating this bug for tracking purposes