Comment 6 for bug 1393391

Miguel: I think we're running into this exact same problem. Can you give me some details as to what specifically the issue was and how you addressed it?

From what I can tell, it looks as though what may have happened on our side is as follows:

* [Event] happened on the rabbit node where the queue was originally created (and thus the clients were consuming it)
* Channels were cancelled
* Because auto-delete is true, and expiry wasn't set, the fanout queues got deleted (because nothing is consuming the queue now)
* Clients reconnected to a new node
* Clients don't recreate the fanout queues (likely due to how it reconnects from a channel closure vs a connection closure)
* Clients start tossing the 404 error and never stop, unless they're forced to reconnect, which recreates the queues

So it seems like your hack is pretty spot on... when the failover happens the queue isn't properly recreated if it doesn't exist, which your code appears to fix.