On AI FM, AFAIK they use SFM - that is, table flight models. Think FC2 player flight models.
Doing a "low-fi" AFM isn't as simple as it might sound, but indeed I think Yo-Yo would have to be consulted about that. I know how it works better than most, but that doesn't exactly mean I'm an expert. Far from it.
And the point regarding models is that a 3D model isn't a "starting point". Think of it like this: do you start design work on a first person shooter with the automatic carbine? Or do you first get an engine that is capable of having bullets fly through the air in the first place? This is a bit oversimplified, but as far as FM goes, the more detailed the simulation, the closer your problems are likely to mirror the problems you have in the real world. And as you might recall, before Chuck flew there were actually serious doubts about whether controlled flight past the transonic was even theoretically possible. It too a lot of testing and a lot of development to get there.
...but once the "trick" was found, things happened pretty fast. (No pun intended.
In an old "table-style" flight model, you can basically just pick up some data from some charts and plug them in. All set. But with an AFM like in DCS, you are truly simulating all of it, and this is non-trivial. This is things that until relatively recently required supercomputers to do at all. ED is doing it in real time, on a consumer PC, AND require it to give realistic results. "Non-trivial."
Basically, what might look like a small thing to an "outsider" might be 90% of the work... Which isn't a dig, it's just the fact that the outsider only sees the things that meet the eye - they see the 3D models, they see the graphical bling, the see the interface of the radar screen (but doesn't see the magic that happens behind the screen) etcetera.
...and the deal with something like a P-51 is that you don't need radars (well, the tail-radar is modeled, but that's really just a glorified proximity fuze :P ), no computer systems, you don't need to deal with the transonic etcetera etcetera. Here the "new" thing is piston engine simulations (which is done at a detail that I suspect will make A2A quite jealous, btw :P ). But for a lot of things the underlying technology is already there. But for a pointy nose, there needs to be a lot of groundwork first, same way there had to be basic tech done to make DCS Ka-50, but after the Ka-50 is here, basic technology for other helicopters is already produced. (Though making other helicopters is still a big job - but the "first" of anything is always the biggest job. And the consumer rarely gets exposed to what the "problem" is since the consumer doesn't see what's going on "under the hood".)