Client based aggregators make me a little nervous because I always think of Primal and their weird cache server model. Obviously, it could be made a lot smarter and less obstructive, but I think doing that on the client level also negates some of an aggregator's benefit, namely the reduction in data use. If the client is doing it, then all the relays being aggregated are still being downloaded per device. Just some initial thoughts, but I think it's valuable to query these ideas, and brainstorm them because that's how we arrive at great solutions!