Reliable one-to-many communication is frequently prone to two serious problems in particular: (1) how to prevent a slow consumer from holding up the rest of the system, and (2) how to prevent massive amounts of negative acknowledgement (NACK) traffic from swamping the network. These problems are related to one another: both deal with the way in which a communications stack (network protocols combined with a middleware on top of them) maintains reliability across a logical network topology with broad fan-out.
