When considering the schema rework, we should consider the denormalization idea.  Basically denormalization is a bad idea so long as we have persistent, important user data in the database.  But it seems like most of our DB functionality isn't in this category to begin with.  If we can move the important user data out of the DB, it becomes an cache/index of the library, which certainly could be denormalized for speed and erased/rebuilt at will without affecting the user in the long run.