I'm not sure what the right thing to do, from our point of view they are 2 different clusters with their own classes. Are you suggesting that when you have 2 different clusters with the same class name, we should ignore the duplication when one is a subcluster of the other? At the moment, we prefer to generate an error and let the user fix its configuration file as it is easy to do so and by making things more explicit the user can exactly control the behavior he wants.