First, before I start with what is (hopefully) constructive criticism, let me say that I really like the overall design. I think the conceptual work is first rate and it's clear that Kael has a lot of backstory stuff that is yet to be incorporated. I also think that a lot of the play effects, like disease spreading and summoning units, are well done.
That said, I have a few issues but he biggest one is the AIs. I don't think they are up to the challenge of the radically different enviornment.
I play normally on Prince/Monarch with a wide range of maps and size types. I usually win on prince and I stay competitive on Monarch. Last night I played a Fall game on Prince and dominated. Not only did I dominate, I took 2 stacks of 4 units each and completely swept 2 opposing civs off the map. My biggest problem was producing follow-on units fast enough to garrison the cities I was capturing. I was also behind my first victim on technology when I started the rampage. The civs never did react like they were at war.
Basically, it seems like they are simply incapable of judging the relative value of all their new options or optimizing at all.
I suspect that the issue probably has a lot to do with the interactions between the large number of prerequisites and allowable units. In the core game, the substantially lower number of restrictions means that it's easier for an AI to optimize as it doesn't have to plan ahead 3 steps to get to a good unit. It just pounds down the tech chart until it's available and if the "random" production doesn't produce a particular building in the meantime, that doesn't change whether the better units become available.
Even in the core game, the AIs are known for not using buildings in an optimal fashion. In Fall, both the large number of buildings with (imho) questionable value and the large number of building prerequisites means that the AIs seem to do especially poorly.
I suppose that on the really high difficulties the massive production discounts might make up for that, but I'm not sure. I have a feeling that most would just end up with a large horde of lousy units and buildings.
I would suggest trying to reduce the heavy reliance on buildings and also make certain that all the buildings have a substantial value as the best ways to increase AI play. There might be other issues exposed after that (like AI use of civics), but if they exist they are currently second-order issues.
===
Again, I want to emphasize that I really like the conceptual work and overall I think this is a good job.