Yeah, agreed. For now the reasoning has been that there is the copy for public consumption, and one for restoring that copy. In addition the same pairing of data and availability exists in another cluster. However, the queries currently do not span clusters in order to locate available copies.