For JGroups 2.5 we have used numerous instances of ThreadPoolExecutor(s) as we needed them. For instance, STREAMING_STATE_TRANSFER, Multiplexer, TP, and ProtocolStack. Undoubtedly this lead to increased overhead and performance impact that we have not measured. First of all we should stop indiscriminately introducing new thread pools as we go further. Second, we should Investigate possibility to consolidate thread pools so they can be reused by various protocols and components of a channel.