For a DnD game, I usually expect to be fighting gods to save the world by the end of a story, or something on that scale.
That’s fair. DnD’s scope makes a single city an awkward setting. (Though maybe Elturel during its descent to Avernus could be a good candidate).
DnD’s focus on combat is also a problem. Either the RPG’s mechanics could encourage non-violent progression through the story, or the story could justify violence, but murder hobos shouldn’t be viable. The selling point of a city is its vibrancy and liveliness; the game should be about options, not combat.
DnD does have cool features that could be put to great use in a city. What if mage hand could pick up objects and pick pockets? Or if Detect Thoughts could reveal basic emotions of passers-by?
As Gray Ghost points out, the third act should bring closure rather than just more stuff. Remaining in a single evolving location could help acheive that, as would maintaining a consistent cast of characters.
I’d also like to see two sets of permutations: One for side questy stuff and one for story. The idea is to increase replay value by having the city include several factions (say 3) and let the game decide for each run which is the antagonist, which is helpful and which seems helpful but betrays in act 3.
I’m just pitching a whole new game at this point. In conclusion, I agree with you it might not work the best for BG4, but an RPG where the point is to navigate the complexity of a big city through charm, intimidation or trickery sounds awesome to me.