Cycling of outbound connections on node restart is suspect and racy

basically Server A can enqueue a "close sentinel" into the queue and then a normal message is enqueued so Server A closes the connection and then immediately re-opens it
the reconnection logic might race with the closing logic from what I can tell