Here's a tech demo of what I've been working on: https://www.youtube.com/watch?v=7q0l87hwmkI
I created a path finding algorithm that can simultaneously path 300,000 units to random destinations at a comfortable frame rate. Units can choose from any of the shortest paths between two points (there are many in a grid), and from those paths, can also choose the path that matches any preferences they have.
Very early stages of development still!
I'll be using old.reddit.com/r/Archapolis if you want to know when the first release is out.
(I don't have an opinion either way, just curious.)
At a country scale some simulation techniques need to change. For example, tilemaps become ridiculously inefficient (a byte per 10m^2 becomes tens of GBs), so they either need some form of compression, or the simulation has to use vector-based maps instead (more like Cities Skylines).
Another quirk is that at a country scale agent-based simulation becomes less interesting, because individual agents don't influence much, only their collective behavior is big enough to matter, and that starts looking just like a normal distribution of the simulation data you put in.
This is very untrue, which is why this problem is infeasible.
It's sort of like simulating every atom of an object vs using Newtonian physics. There is a difference in accuracy, but it may not even become apparent or matter for gameplay.