OK, there's one change in game the massing causes: at some point, it's impossible to attack due to economic effect of too many units. At which point there should be no more war (it's cold war, everyone has too good defenses for anyone to attack) and thus the one with most land (and best economy) wins when everyone else gets cobbled by the expenses.
Elandal hit the nail right on the head and my 3 recent games all back that up. In my latest, I share a continent with Toku (playing as Ottomans). We have a Mexican standoff for while and we split the continent about 50/50. In the late Classical, he DoW's me. We fight a relatively inconclusive war (although 50+ units are killed on both sides). We have a period of piece and buildup and then he attacks again in late Medieval (although I know he'll probably do it so I have built up too). This time, I manage to take one of his cities by the end of the mess and again, we both lose copious amounts of units.
Somewhere in the middle of our war, contact is made with the other continent. One of the AIs (Hatty) is a vassal to the Vikings. Another AI is alone on their own continent. And the last AI is (Vicky) is best buddies with Ragnar due to religion. The result? All of the other AIs (except Hatty) have nearly double the points of Toku and I and are miles ahead on the tech tree. The reason: they didnt go into the ridiculously expensive 'death spiral' of going to extended warfare.
Elandal is indeed correct and it was something I had sensed, but never really put the finger on. When the numbers of units are as big as they are, the cost for going to war is IMMENSE. Even a 'successful' war will often leave you behind others who are at peace (but still building up their massed defenses). When stacks of 20+ are clashing, you could be lossing 1000s of hammers of production
in a single turn!! And you MUST replace those units or else be weak and eventually overrun.
It goes back to what I first stated...the massed militaries cost so much that they mute the other effects in the game. And in the early game, you might gain some success relative to the cost of those troops by attacking a neighbor. But from Medieval on, the cost of going to war is just crippling. And thats why the more warlike AIs always tend to peter out and fizzle...they come out strong and may take out a rival or so, but then they hit that wall of expense vs payoff with the huge militaries and they crash.
If you happen to be one of those involved in the early war with a warlike AI, you'll struggle to remain alive...there is no doubt that the AIs can prosecute wars much better now, but the cost is astronomical. But its likely you'll never recover from the early wars (assuming you win). And on the flip side, if you manage to stay at peace (through diplomacy or simply building a bigger stick), then you'll likely cruise to victory because the AIs will either destroy themselves and then fizzle or else poorly balance econ vs military and also fall behind.
I also think that Elandal is again correct on the consequences of having the AI not build as many units. He is correct that yep, they may lose a war to an aggressive human player. But thats fine. As long as the cost of prosecuting that war is balanced to the gains, there is NO PROBLEM. I agree that in 2.08, the cost/benefit ratio of attacking many of the AIs is very lopsided in favor of attacking (Blake's 'pinata' effect). But 1/25 is the polar opposite. It doesnt need to be anywhere NEAR that high. As long as the AI makes a 'credible' defense, its fine. 5-6 units per city will indeed lose to human stacks of 15. But that doesnt mean that the AI should be building 15-20 to resist. That just encourages the human to build 20-25 and the cycle begins again, only with the cost of warfare overshadowing all else.
At some point, it just has to be accepted that the 'better' idea is to simply make the cost of an attack the deterrent. And I believe that 5-6 units plus some reserves would do that. It would no longer be the cakewalk that it can be in 2.08, but a real cost will still be assessed (15-20 units still isnt cheap when the rest of the world is teching/building infrastructure).
So, FWIW, my opinion is that the AIs should build perhaps 2x-2.5x their 2.08 levels. If at war or actively building to attack, that should increase obviously, but for 'peacetime', the AI should cap their unit production and continue to develop. If one or more is killed off, so be it...the rest will still be presenting a challenge by continuing to progress.