imo it is a mistake to work on scaling by messing with the protocol. getting creative with the implementation of mining node is the way to go; it is single-threaded right now #facepalm
a competition of different creative node-scaling implementations is what is going to bring huge-block-support; stable protocol allows paralellization of these implementations
if node-devs of all implementations are busy aligning their impl with the latest protocol-changes, they aren't spending time on eliminating processing bottle-necks in the node-software
a miner can run an open-souce node impl that scales the best for his setup or invest more on a proprietary impl for additional advantage
though this is only really going to become a factor when the blocks get to ridiculous sizes
