There were substantial problems in the logic of PostInitializationCommunication when I originally wrote it up that should be resolved by #118. However, we probably aren't testing this well right now -- add a unit test that initializes a basic problem?
