Yes! I made a comparison here arguing that multi-tries are the way forward. Do you agree with my assessment?
Absolutely. The last two ideas in my original post (partial statelessness and 224-bit hashes) I added only for completeness. I do think that multi-tries are probably a simple and worthwhile thing to do, because they can provide a ~3x reduction in witness size, and have the additional parallelism benefits.
In general, I think it is a very worthy endeavour to have somewhat “fat” shards with generous gas limits. The reason is that I expect cross-shard communication to be at least an order of magnitude more expensive than intra-shard communication. We want related to applications to fit in a single shard, and we want to avoid splitting large applications into too many shards.