Maybe 12 is too low. Should be 15 or 20... Anyway for balance reasons, a large city should be much more difficult to conquer, while a 1 pop city should be really easier. It is already because of the culture effect, but having a larger number of defenders would make sense.
A larger city has a larger production capacity and can replace losses easier. Unless its a border city (in which case you've should've had units on the border anyway), an approaching army usually has to march through the country and if if you have defender units in hills, forests, and forts, then an approaching army
should be shredded before the army reaches your larger city.
Also, the defender bonus+defensive buildings should be more than enough. After all, a well-defended city with powerful defensive units, even at the limit, can take a while to crack (I know this from playing the feature in another mod) unless you bring A.) Siege units and B.) Surround cities and then C.) Cut the city off from the rest of the country. In any case, the enemy would have to surround your normally high-production city (which I assume is true for most high pop cities) or have a pretty mighty tech lead on you to break the city open on one blow and you usually have time to send a relief.
While there should probably be more units in a city than on the field at a time (probably, I personally don't like that part), the number of units that are allowed to be garrisoned in the city should be static across the board.
Which stops turtling and fantasy Stalingrads.
On the other hand, terrain
seriously affects invasion plans. For instance, today I was going to take a barb city at the tip of a peninsula but there's a mountain range that blocks a land route to it so I had to land forces on the only two possible tiles in that I could land on (accidentally landed on spearmen in forests which took out a ships worth of war elephants and horse archers) but I landed and since there
was a tile limit, I could only bring to bear a certain amount of arms on the city which had archers and spearmen.
Long story short, took a few more turns to take down the city that didn't even have a cultural defensive modifer or terrain advantage (built on grasslands!).
Good point. Does the limit apply to stack of units at the end of the turn, or should the traversal of a complete stack be impossible? I think the first solution should be better.
The former. Say that you have a stack of 3 and a stack of 5 and the limit is 6. When you try to move the 3 through the 5, only one will go through at a time.