[rabbitmq-discuss] Consumers from slave queues on different nodes and duplicate messages

[Dave Stevens]

When multiple consumers are consuming from a single, logical, HA queue
mirrored across N nodes, is there any potential for duplicate messages
beyond the documented case when a slave is promoted to master?

I am thinking specifically about the case when consumer A is consuming
from slave queue on node X and consumer B is consuming from slave
queue on node Y. From very superficial testing with noAck=true, it
appears as though the master queue only delivers a single msg to
either A or B as desired, yet I'm assuming it must notify the other
queue of the msgs existence. Is this in fact correct, or must I be
concerned with the msg arriving at BOTH A and B. Does this behavior
change when noAck=false since master is no longer Acking the message?