I've just used a classic name for the genre. I'm a big fan of Cities: Skylines.
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.
>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
This is very untrue, which is why this problem is infeasible.
For example rush hour is an emergent phenomenon. But it's something that is happening pretty regularly depending on typical work schedules. You can simulate thousands or millions of agents with their intricate goals of their daily lives to have it emerge naturally (and it's very fun to program that), or you can just hardcode fixed times for rush hours. In a big-picture view of country-wide statistics the difference between these approaches is underwhelmingly small.
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.
The techinical term is "genericised trademark" IIRC. Same goes for "Civilization" upthread, and for things like "Tetris" or (edit: to the extent trademark offices are corrupt enough to register it in the first place) "Chess".
(I don't have an opinion either way, just curious.)