Is there any upgrade to Bitcoin that could make it possible to do this without centralized permissioned servers and the single point of "failure"? Can multiparty computation, sharding, federation, or any other technique be used to split up the private key, remove the need for a single trusted server, and remove the need for the special chip?

To be honest, this is AMAZING work, but without that, I see this as just another in a growing number of "tools in the toolbox," without really being the end-all way to scale.