I think these are the key--just overall, lower the resolution of the combat in both time and space. So instead of 2m hexes and individual segments, go a turn at a time with city block-sized hexes, and work out how many mooks on either side get killed.
This way, if there are supporting-cast combats happening alongside the main PC-villain combat, it can still happen without causing too much delay. Basically each post-12 you would just work out what's happening between the National Guard and the Chitauri, and where the big crowds of innocent bystanders are.