The main issue is real physics is kinda chaotic. So the more liberally you apply it, the more likely the chance is you are going to run into ugly corner-cases that waste dev-time patching up.
At the end of the day, a game is just a type of show. So if a solution comes along that allows you to do specific complex, realistic physical interactions in a somewhat predictable manner then that'll nearly always get the nod over a more generalist but potentially finicky solution.
And particularly when the pop trends are realistic scenarios and naturalistic (or hyperrealism) look to objects and surfaces, the over-the-top trends of making AAA games specifically as kooky physics playgrounds has fallen off. (Except in the places where it is popular, as actual physics playgrounds like TABS, Boneworks, Beseige, Teardown, and Instruments of Destruction, but those don't have to worry about story or realism or really even "balance" where a physics effect gone wrong accidentally breaks forward progression in the game.)
There are titles like Rainbow Six Siege which are trying to blend realism and play-realism with physics. (And I wish it did it better; adding breachable walls is great and R6S does have rules/cheats to maintain map design with not all breakables, but furniture and decorations should be flying... it is a multiplayer game, so that makes it harder and more susceptible to hurt the ping, but physically-activated objects could only be computed locally and clip like heck for all I care, I just want shit to fly off the shelf.) It's not so common though. Much more often, your prebaked physics are fine for selling an exciting moment or a type of big hit, and then everything else that the player wants to shoot with a shotgun (the pillars, the trees, the windows,) the game designer doesn't really want to be shot up with a shotgun because they placed most of that stuff in the scene to establish the map or to close off access beyond their façade.
BTW, DMM works great for what was shown... applying it beyond that may not as easy as you'd think. For instance, that breakable wood concept, that works great if your surface is plywood, a single sheet surface that can splinter along pre-set or algorithmically designed fracture points depending on impact point. You get tensile strength, you get gravity, you get fibers depending upon each other for strength, you get a remainder of an object which can be further acted upon, all good stuff. That's a 2D plane though (with some abstraction, that they then apply some color or texture inside.) Apply that to a round tree and it's not going to work, because trees do not splinter like plywood. What they actually ended up with in the game did some of this, but the DMM implementation in Force Unleashed was nowhere near the complexity of this single sheet (and in-game may have used a lot more traditional methods of physics simulation in cooperation with the DMM definitions.) People see, "Ah ha, we now have
wood in videogames!" and imagine destructible forests and splinterable scaffolding for a building and all that... but what you're seeing is how this piece of wood works, not all wood.
Especially as objects get more complicated, you're lucky if you're running into such simplicity of material here. Somebody else came up with the example of a picture frame, which not 1 thing, it's wood, glass, and paper; each element would need to have different physical properties with their own physics simulation that somehow would also need to collaborate between each other in that one object. Even if your game engine could allow designers could assign "this is glass, it breaks like glass; this is wood, it breaks like wood" (and they
kind of do, although wood seems to be a real blindspot for something like Unreal Chaos Physics when I'm reading up about it, not sure how well other physics solutions are doing in this respect?,) only so much of your scene is that simple, and only so much of even the simple stuff do you want to be activated physically in an unexpected way.