Just something for information purposes...
One way I've been able to 'workaround' this issue is:
1) start up skyrim
2) go to task manager, set up TESV.exe affinity to use one core only (disable all but one core)
3) go back to skyrim, load your save normally. It will not crash
4) once the game has finished loading, go back to task manager, enable rest of the cores
5) back to skyrim, profit.
It may be more simple to just load up a 'janitor save', but this may give a bit more insight into what exactly is causing this issue. Maybe it's something related to thread safety in the game. If there was a way to make TESV run on only single core until game is loaded, that would most likely make the issue go away. Failing that.. I don't know. Would need to find out the actual cause of the instability. Maybe it's loading something animation related on separate thread, and if it hasn't been able to finish that before certain point, it'll crash when the game world fires up. That wouldn't explain why it actually works when loading in 'quiet cell' though - I seem to get this most often when loading in large areas such as outdoors.. while indoors areas are generally ok, even with couple dozen actors in the area.
On one rare occasion I got a save that had same issue but behaved a bit differently: instead of immediate crash, I loaded into game in T-pose.. and game crashed as soon as I tried to do anything that would initiate an animation. Draw weapon, try to move around - anything. This was inside temple of kynareth, so the area was relatively small, but there's still a handful of NPCs around. That save, too, loaded fine if I first disabled all but one core. This seemed to be infinitely and reliably repeatable - that is, every time I loaded that save, it loaded into t-pose and crashed when initiating an animation. Rerunning FNIS didn't help - and since it -did- load fine when disabling cores, it wasn't an issue directly with missing animation or the like.
This is pure speculation, but it's possible that none of the NPCs there were actually using an animation at the time (sickly farmers laying still on beds, priests etc sitting on chairs and such), so it's -possible- that no character at all triggered an animation at game load - and thus it wouldn't crash until I myself started an animation. Just saying that it might support the idea that the actual crash is triggered by the game trying to run an animation, when animations have failed to initialize. Of course the actual problem would still be that animations failed to load.. and as such the fix would also be related to this.