try to resolve race condition, third time

[s074]

The problem was not properly addressed last time because the exception was crashing the whole connection instead of just the channel, so the exception got thrown again when we tried creating a new channel on the old disconnected connection.

To address the instability that is introduced by publishing to a queue that might or might not be there, due to the race condition, we now isolate the connections. We have separate connections for publishing and consuming messages, this way, when the exception triggers our publishing connection to close, our event consumers are unaffected since they will be using a separate connection. This also means that we can easily restart the connection without worrying about having to reconstruct any state (ie recreating all the user channels and restablishing all the event consumers) since the publishing connection does not have any event consumers and was used purely for publishing events

[MathMan05]

How's this going?