the more stuff you have in the game, the harder it is to have good AI and good balance. MoM and MoO II both really captured my imagination with their choices and detail, but the gameplay was always seriously held back by horrible balance and awful AI. Creating a balanced MoM with an intelligent AI would be all that it would really take to dethrone it, but that may be easier said then done. I think this is why people tend to simplify, having more choices isn't necesarily a good thing unless every choice is meaningful and unless the AI can use those choices in an intelligent way.
To add to your post, I think AI should be taken in context as to when the game was created. MOO1/MOM were 93/94 titles... I believe I played them on a 33mhz computer and turns took a couple seconds to process, so there wasn't a lot of computational power being used by the AI and there weren't AI coders that had a decades of experience of strategic game AI to do the work. Contrast that with today's hardware, where with multicore processors you can now have an AI dedicated on a single (or multiple) core running in the background at the 2-3ghz range without being noticed by the user while he/she plays the game. Mind you, I don't think anyone would ever find the code to utilize all that computational power in the near-term, and the devs would need some very skilled individuals on the AI team to begin with. But with grid-based tactical combat, you can now for example have the AI run a lot more permutations like a Chess-AI to better its movement and choice of attack, compared to say MOO/MOM where units generally just blindly charged at you (unless they were designated as ranged fighters).
I think AI is something that should be evolving naturally as time goes on, so it really is a cause for concern when any modern game has worse AI than its predecessors. Elemental should (in time if not already) have better AI than GalCiv2, otherwise I'd be fairly disappointed with Stardocks performance in that regard.
I agree with you that having a ton of options just for the sake of more options isn't a good thing and options should always be about having meaningful choices. But on the other hand, I think a designer should look at a game like MOM for example in TBS, and look at all the features and interactions of systems it has, and then one by one strip them down to what doesn't fit or work for their TBS. But before they remove a particular layer, they need to have good reason to leave out a certain facet of the game and to understand what the impact of removing that is on the rest of the game and have something else to fill in its place. You can really miss significant nuances of a particular game and really simplify it too far.
Case in point:Demigod mentioned earlier. Also, flying units can be another specific example. In MOM flight had uses on the overland map to cross seas or some terrain, while in combat it allowed you to dictate which units you'd attack and when, or sometimes cause the turn limit to expire. You could have other movement spells or abilities on the overland map to replace flight in that regard. As for combat, all you really need the AI to do is be smart enough to cast a disenchant flight, or some sort of grounding spell, and flight doesn't become so useful anymore. Or with Stardocks special unit abilities those could be used instead, maybe spiders could be a partial anti-air unit by having a spit-web ability (or whatever else works thematically) that grounds fliers that can then become surrounded and beat down while entangled (thus giving the spider side the ability to take away initiative in commencing combat). If flight isn't in the game though, Elemental will have to rely on other depth that will hopefully be able to make up for its absence. Removal of flight in itself is insignificant, but if dozens of aspects are removed, they start to add up.