If Approach 1 is adopted, how much of the committee & subnet sharding could be simplified / removed from the current protocol?
My understand is that both the current protocol and full DVT approach are topologically very similar - the main difference being that DVT has the potential to reduce load by taking the validator sharding and signature aggregation schemes out of protocol.
Giving individual operators the responsibility to form there own committees naturally makes the network more performant. For example, the current protocol may assign two validators on opposing ends of the world to the same subnet. This problem is resolved when operators can manually join a DVT pool with peers they know they can perform well with.