Almost so much that it becomes easy to ponder if we're already in someone's 3D simulation ;)
Ray-traced voxel based worlds. Personally, I think voxels will be hard to manage for objects/characters. Someone smarter than me will have to solve that problem.
I believe the appropriate expression is 'Horses for courses'.
A classic way of solving the voxel vs. smooth object problem: http://en.wikipedia.org/wiki/Marching_cubes
Someone smarter, or someone with specific knowledge of the relevant domain and a lot more time to invest on the problem? I see a lot of people assuming that someone who can do something they can't is inherently smarter, even if they do something just as complicated for their daily job and think nothing of it.
The original Red Faction had largely destructible level geometry as well. However, that was just based on explosions deforming the terrain, rather than a physics simulation of destruction like RF:G. Polygons would fall if they were attached to nothing. The multiplayer level set in the natural canyon/bridge area is a good example of this: the large bridge in the second level (3 paths) is able to be taken out with some rockets and persistence to actually fall into the pit in the center. However, there was a limit to how much the level geometry could be deformed, and after a certain amount, explosions would stop removing sections from the terrain. It takes a good amount of time to get to this point, but it is possible. Also, the developers did a great job of hiding some interesting secrets in the deformable terrain. The multiplayer level with the two buildings on opposite sides and multiple levels with the desert in between has a hidden fusion rocket launcher (BFG, essentially) on the roof that's only accessible by destroying the terrain to get up there if my memory serves correctly. Also, the lobby level has four sections hidden behind walls, and a hidden roof area. Clearly, they put some work into using this feature well.
In Red Faction: Guerilla, only the buildings/objects are destructible, but there's more physics modeling. Taking out key supports will bring down buildings...in theory. I'm not a structural engineer, so I can't speak for its accuracy, but it often seems as if buildings are much hardier than they should be. Still, it's rather cathartic to take down an entire enemy barracks with a few explosives and a sledgehammer.
As for the actual math and methods behind this, I cannot say. I just know how it worked in gameplay from many hours playing the old RF multiplayer, and some recent time put into R:FG. Feel free to ask if you'd like to know more though.
(Huh, I never thought my videogame knowledge would be all that relevant here.)
This simplicity makes reasoning and manipulating the game world less challenging, resulting in the game being more accessible to a wide variety of players.
I've had to switch languages a lot when coding, but even in a constrained or flawed language, you see code that is beautiful and amazing. Constraints free up thought, as when you stop rebelling against a framework, you can excel in it. It's hard to push your brain into what feels like a hole; I can't get anything done in Perl because I spend too much time being repulsed, but some people embrace the language and produce great software.
Having just praised restrictions like that, though, I still have to mention that this game looks really cool.