A failure of a node in our main message queue cluster prevented messages from being processed until the node was brought back online