The idea is to expose more of the internals of the clustering setup:
And start exposing some features of JGroups, like automatically choosing a leader among the members.