Fixed bug where we might drop some fault events that should be delivered when we are not
able to connect. We batch up messages waiting for a clientId to be assigned in the
clientIdWaitQueue. When flushing this queue, if we have no clientId (i.e. we are likely
faulting these messages), if there was more than one message in the queue, we'd just end up
putting the messages right back in the queue instead of faulting it. With this change,
we null out the queue temporarily so that internalSend faults the message.
blazeds checkintests: pass
Reviewed by: Seth