At the moment, it is very easy for an application to start a huge number of putAsync operations and fill the transport executor's thread pool, delaying internal work such as state transfer. Increasing the size of the transport executor's thread pool won't work, because that would in turn fill the remote commands and JGroups' OOB thread pools, with the same effect.

If the cache async operations used a different thread pool, it would be possible to configure more remoteCommandsThreadPool threads than asyncOperationsThreadPool threads, avoiding this problem.