It has to be one of the most robust meta-feedback loops, because the "vision" for the scene is apparent to everyone in real life.