Aside from issues like defending your cities and the time and resources to build them, I think the mechanical issue here is prestige - you get a bonus to growth in your cities equal to your prestige divided by the number of cities you have.  So if you have 10 prestige and 40 cities, that's only one person every four turns in each city, but if you are down to only one city that's 10 people per turn...

Personally, I think that having control of the land and the resources is a bigger help than the slow base growth on cities is a hinderance. So I often have a lot of cities.

But that brings up another issue: if you have a lot of cities and no challenges, the game gets boring, and it's probably time to end that game (I prefer diplomatic victory, here).

You can get around the slow growth with life magic (Sovereign's Call) or with Outposts (I forget which upgrade, but it gives +1 growth).  It is also a large incentive to have a high-level Sovereign, which leads you towards exp bonuses.

There doesn't seem to be any downsides with building as many cities as possible, aside from the headache of trying to micromanage them all. As there's no autobuild option for cities you need to babysit them, but assuming you babysit them and build the proper improvements in the right order, along with plentiful outposts for consulates, you can fill up huge amounts of land with highly productive cities.

I usually build fortresses outside of my core. The core is towns for food production, then around that a ring of fortresses. And then everything outside of that is fortresses. Fortresses are mostly hands off assuming they're not directly on your front lines.

More cities also gets you a bigger industrial, research, and economic base. More taxes raised, more research done, and the ability to produce large numbers of units at once should you need a zerg rush.

I like to play the game with world difficulty set to insane, monster lairs set to dense, and game speed set to epic(slow research/build).

One of the side effects is that i do not have many cities simply because i have to wrestle those good spots from insane monster armies. The sovereign at start can barely handle mites(i play stock gilden, not custom) and only a few squares into the wild the world is populated by scary 120-hitpoint black widows , then 800-hitpoints troll-warrior armies etc. Just making the one starting city survive can be a challenge.

The growth really is a non-issue for me most of the time, since the cities reach food cap alot faster than i am able to research/build more food-producing facilities, so more often than not the cities are sitting at max population waiting for the tech to catch up. Only exception is when i found a new city when the research is all done already, which usually means mid to late game and the victor is more or less decided already.

Thinking about this, the choices really depend on the kind of game you have been playing.  Personally, I have been playing in a fashion similar to stax77, and I am going to walk through some of my reasons:

I play on epic pacing since the game has not been balanced for late game research.

I play on a huge map since various bugs in 1.02 give the AI an unreasonable mid-game advantage.  But this makes the AI too weak.

I play Pariden on insane difficulty since the current AI is no challenge on a huge map.

I play sparse resources and sparse magic since otherwise Pariden would be too powerful.

I play dense monsters (and insane map difficulty) so that combat is not too easy.

I play only empire opponents to ensure that they will all declare war on me.

I only use the four "stock" opponents because I will be far too weak to withstand an early game rush.

I use frequent random events to try to make the game challenging again.

I use no champions, because otherwise Pariden would be too easy.

I use a custom Procipinee because the stock version is uncomfortable in this environment.

The result is a game that is big enough that I can survive the occasional <<unit running off into a monster because of a UI bug>> and with enough going on that I can ignore the AI until I discover them, and with an AI which is not a complete pushover when I do reach them. It's easy, relaxing, and not without challenges. (Edit: unless I land in Yithril's back yard before I can prepare myself, ouch.  I have a shallow initial growth curve here, and am hoping to get myself above the flatter (but initially steeper) growth curves of the AI before I meet them), and epic jug armies are hard to deal with when my best army is mid-level. 

In other words, I am playing in a fashion which makes the AI nearly useless, and then using game difficulty settings to make the game challenging.  Some of the things I am doing to make the game challenging for me also make it challenging for the AI. But, ultimately, the above choices are based on the strengths and weaknesses of the game in 1.02. And, anyways, the kind of game that results is just very different from the kind of game you get on default settings.  Also, playing pariden in a sparse magic world, I must expand heavily because that's the only way to build up a mana reserve sufficient to deal with the game challenges.

(By the way: anyone playing against insane ai difficulty is taking advantage of defects in the AI, from an "end game seeking" point of view.)

Anyways, the point of all of this... to intelligently choose between your research and production options you need to have a roughly accurate idea of how long the game is going to be.  Take "Refined Research" for example -- this gives you nothing and only becomes valuable if the time remaining in the game when you research it is greater than 10x the time you need to do the research (the exact value here changes slowly if you repeat this option).  Something similar happens with production and growth -- you need to have a roughly valid estimate for when you are going to need the growth or production to determine the relative priority of that when compared with other options.  The challenge of the game is when the answers include "now" (or immediate future) along with "later" giving you no particularly good choices.

So when looking for advice about choices, I think we should keep in mind that specific answers are tailored to specific situations, and the value of that kind of answer is based on either {a} how close the situation the answer is based on matches the ones you will be facing, or {b} how well the answerer describes the issues which will matter to you when you need to make the choice.  [And, hopefully without going into excessive detail, like I have done here.]

Edit: it's looking like my earlier success might have been luck -- I think the combination of insane with pariden with scarce magic is just too much for me, and I will need to either use a lower difficulty or normal amounts of magic. I hope 1.1 gets released soon...