Failure to receive state from the coordinator does not prevent starting a cache. The HAPartitionImpl that's starting just assumes it is the coordinator and logs an INFO level message.

We need to determine for sure whether we are the coordinator, and if not, throw an Exception.

We also do nothing to ensure that the setState() call (which is happening in another thread) executes successfully. We need to look at how TreeCache handles this as an example of what to do.