As with many things in design, elegance is served with regularity. If we design a mechanic so that it is like a host of systems, each of which is a version of the same thing, then the game
feels more cohesive no not that word, but... well it will feel simpler but also extensive in meaning. So, for example, returning to the idea of connectedness being a root of gameplay, the city would have its allegiance, which makes it your city and not another player's. But it also has a majority religion, right? And it has other religions in it, but one of them decides or controls what 'religion stuff' that city does. The institution of a policy, from the city level (player) could make it so that religious tolerance is in effect and that the effects aren't of a domination variety where one religion rules and the others submit.
A city could have a composition of -any number- of these substances. Substances like religious adherence. The sources , coming from other populations (settlements), would influence the target city through the connectedness features, contributing a magnitude of "influence" , the meaning of which depends entirely on what the simulated substance is. And it shifts the local situation. But a rule decides what state the city is in from the measurements of those influence, -within that subsystem-, meaning, it just selects the contributing influences and assigns a rank to them, possibly starting out as just 0 (not in majority) or 1 (majority).
Cities being settled inside your region is extremely annoying because it's frustrating, and also stupid, and looks like an exploit, and also stupid. Settling colonies is something we want to see, while balancing how there are some difficulties which should appear while not also being utterly impossible. Connectedness is, once again, a strong device for cutting up the scenarios along the lines we want. Maybe, if you don't fully enclose a space.... and another people maintain a road , and have a very strong identity of religion, or perhaps a civil rights concept like citizenhood, then they will persist in their consciousness as distinct from the surrounding state. And success in that distinctness, too.
I'm with Boris on the idea that going into simulating pops will multiply and create headaches, but, perhaps we could do the heavy work for the player, list out what elements we sorta roughly kinda want to be a factor, and then make a simple cutoff rule that keeps the actual game in a state impacted by the simulated concept. Just enough , of things we design to contribute, so that we and the player can look at it and go "yeah, that's immigration." We don't make a formula of any complexity, and we certainly don't hide math you -need- to know inside the manual. The player needs to know what next turn will be, and we just put an immigration term in the things it affects (e.g. just population and religious devotees) and the player can expect it to be equal everywhere it exists and the determination of if it's there is no more than any other math you do for, like, the exact size of a food bucket or, at most, for a combat damage spread.
My point being, first, let's imagine it more ideally than the game will actually be. Then corrupt the hell out of it to cram it in to the small head space. We can optimize the distortions only going in that direction.