Possible isolated consumer with OpenWire

It appears that there is a possible race condition with OpenWire in which a consumer can be created on a connection that is closed / closing. A recent investigation of a heap dump from such a case revealed a consumer on a transport that was not yet itself marked as closed, but the underlying socket was in TCP_CLOSE and the channel has closeInitiated set to true. The end result is the possibility of infight messages for a consumer that cannot be removed without a broker restart.