||Distributed Enterprise Messaging with MantaRay|
Ok, that QueueCoordinator was used due to the flip-flop problem and due to the Queue semantics.
However that QueueCoordinator seems to me just like a "mini-broker". It can be run on top a producer, consumer or in between, yet it remains a centralized node.
Such a concept doesn't seem to justify the promises in the article - "fully distributed", "No Single Point of Failure", "No Single Point of Congestion", ...