Haven't read all posts here, but I don't see the issue, really.
There has been a lot of talk regarding the "standard start" thing (you know, starting the game in the same way every time, like researching the same stuff in the same order because of some ultimate progression).
The "complicate it still further" thing really reduces the risk of these "standard starts", because if there are only a few special food tiles in the world, then you'll have to consider whether you want to wander aound and explore until you find a food resource, or if you'll simply settle for a "garden-city" and look for resources later. It depends on whether you get a lucky starting position, and in luck, there's no "standard". It would be different every time.
Now, I'm not saying that only building gardens is any fun, but by adding more kinds of improvements, and, like Frogboy mentioned in a Dev Journal, being able to merge improvements of similar types when putting them close to each other, there would be so many possibilities that worrying about gardens wouldn't really be an issue. Smart city planning would really become key, and cities would be more realistic.
Say, perhaps, that you place a command post, barracks, an archery range and an armory next to each other, and it "morphs" into a "military academy" with stats equaling the former buildings plus bonuses, for example. Imagine how rich the city building would be, almost like a crafting system! And to return to the garden thing, gardens could gain bonuses when placed next to each other, perhaps in a 2x2 set they could become "farmlands" or something, which would have bonuses.
But the most important aspect to this, I think, is that it could remove the need to delete outdated buildings. (The following is just an example for examples' sake, I know it might not be accurate in actual game rules:) If I had four merchant improvements, for example, for a total of 4 gildars per turn covering 2x2, and then researched the market improvement, which is also 2x2 in total size but gives you 6 g per turn, I'd normally have to remove the merchants and replace them with the market for optimal gildar per used tile, which is boring and "micromanagementy". Now, what if the system worked in such a way that the four merchants had merged into a "merchants' quarter" which gives 5 g, and then, when I research the market tech, the market replaces the merchants' quarters giving 6 g instead. Instant optimal upgrade without having to replace anything manually. It might sound confusing written like this, but essentially it would be: 4 g -> 5 g -> 6 g, without having to micro and remove old-tech buildings and build new ones, while still not sacrificing complexity (eg. not just having one single economy improvement that upgrades through three steps, for example).
Now this can get very complicated. But, a nice way to solve that would be colors. You know, at the moment, all the buildable squares are green. But with a system like this, it could work like this instead (example):
I have built a commandpost. Now I have researched barracks, and I'm going to build one. All the tiles around the commandpost are green, because placing the barracks there would build towards a military academy. The other squares are yellow, because placing it there would not build towards any bonuses.
To further simplify things, there could be a note in the improvement description, for example: "Builds towards: Military Academy". The ultimate thing would be to make the words "military academy" clickable, which would lead to a little popup with descriptions on how to build it (unresearched improvements would have to be listed as unknown or something, though), for example: "A military academy consists of: Commandpost, UNKNOWN TECH, Archery Range, Barracks. *stat summary follows with bonuses made clear*".
So, yeah, that's what I had to say. I brainstormed all this just now so it has some rough spots, but I believe it would really make for some rewarding city building and really separate the "men from the mice" without being hyper-micro or time demanding, instead just relying on city building knowledge.
EDIT: I realize now that this could work against snaking (building a line of 1x1 improvements to reach a resource), because all those 1x1 buildings would miss out on bonuses.