Yep, in a nutshell, it's just a lot of work, and some might say the gain is very little, unless the game is built around it or acts as a demo to show off technology (hello, crysis).
It's also puts additional strain on the engine and hardware, not to mention debugging the whole thing in the end.
Static models are much less work, you make them, set the lighting, and you have yourself a complete, rendered scene. When you start tearing them up, you suddenly need to calculate new lights and shadows as the destruction progresses, get new assets to reflect the devastation (holes, debris, chunks). Not to mention, coding all the physics to make it all fall apart in a believable manner.
There are also other things, like teaching the AI to cope with changing environment (I am not necessarily saying: use it in a meaningful way, basic stuff like making sure that it doesn't go tits up when a wall it wanted to use is no longer there - it's so much simpler with static environments).
The manpower needed is probably beyond the reach of indie developers, unless we're talking autistic savant level of committment. Though in my experience, guys capable of writing such a game engine would rather put in on sale than make their own games.
Possibly current AAA titles would have the means to make such systems more popular, but AAA development is very risk adverse and would rather blow that money on marketing, voice actors or any such shit.
Not to mention, your average console and at home calculator PC wouldn't handle a complex destructible engine, and that's plenty of lost sales, my good man.