ElasticSearch just handles split-brain differently, by shutting down nodes that are cut off the cluster and marking them as such. The problem is present in the _default_ configuration for more than 2 nodes, but ES is very well tunable to your specific cluster layout (which you should do anyways).