It doesn't seem like the game is that taxing on decently good processors. It seems like you should easily be able to do those calculations if you have some extra cores lying around.
One thing to keep in mind is that even in big, 16 city regions, only subregions of 4 cities each are actually connected by roads and rail. In fact, these subregions are practically autonomous, except maybe for air travel (I don't know). This is why you have 4 great works per 16 city region. So first thing to do would be, if there are actual calculations taking place between these subregions, cut that off.
Now we only have the interconnections of 4 cities to deal with. Only 1 of these cities is 'active' at a time (the one you're currently playing on). I don't see why you couldn't do a rougher simulation of what is happening between your active city and the three other cities. You don't have to track agents from other cities to the active one, just have a counter that keeps track of total population.
I may just be a naïve computer science student, but this doesn't seem terribly hard to do if you already have everything else in place. And it couldn't possibly be slower than what they currently have: I've watched one streamer have 3 cities within a subregion all have different values for progress on a great work. They weren't synced up at all. It was just silly. On one local machine, there's no way this could happen.