I'm wondering if there might be some sort of contradiction between intelligent end-points and server side computation. That is to say would a focus on intelligent end-points discourage the use of server side resources to pre-compute information relevant to the clients in a manner that saves overall resources. Taken to the radical extreme aspiring to the virtue of intelligent end-points might push one to relatively inefficient distributed systems.
I also wonder if some insight might be gained by seeking to generalize the proposition beyond computing systems, and shifting focus from the objects to the relations. For example perhaps communications in a business should maximize information density by crafting communications protocols which reduce book keeping cost.
Perhaps analogies from this domain could then be transferred back to the original domain, for example what can be concluded about the reduced book keeping cost of universal in contrast to means-tested benefits. Perhaps that our compound data types should be standardized, and that protocols should be established on this basis.