Currently, JGroups transport thread pools use discard as a rejection policy. Thus if there are no available threads, and the queue is full, the executor will silently drop the task.
Abort will throw a RejectedExecutionException, in which case, which JGroups catches to populate the num_rejected_msgs statistic. This is a better default, so that users can monitor this value to determine whether or not to modify their thread pool configuration. Otherwise, num_rejected_msgs is 0, which might be a source of confusion.