I'd say the intense study of independent agents arises because fast, scalable systems necessarily require systems running in parallel with tolerance for failures. We design systems like that only because it's necessary, not because it brings benefits other than speed and scalability.