This seems vastly incorrect. We get these super-realistic graphics because CPU power has advanced AND we still use those optimization strategies. Games with vast open environments apply model swaps for far-away models, and we still subdivide the world to speed up rendering, so no, we haven't started taking shortcuts; this stuff is still as complicated as ever.
Also as mooseodeath said, it doesn't matter how awesome it is. It is infinitely impractical to apply this sort of thing. It raises all sorts of technical AND design issues, all for one cheap "WHOAAAA!" from the player.
Also, HL2: Episode 2 had completely prescripted destruction. Even the exact physics of those explosions was precalculated, so it's essentially just an animation.
It's quite the simple (and sad) equation: Graphics sell games. Gameplay doesn't.
You can print graphic screens on the box. You can send out screenshots that wow people. You can make cool-looking trailers for YouTube and Gametrailers. But all you can do about gameplay is list features that may or may not be actually included - how many games state "never seen before interactivity" or "the most thrilling gameplay you have ever played"?
Great gameplay makes people want to play the game again. AFTER they bought it - so from a business and marketing perspective, going for stunning graphics is always the better way.
Destructible environments certainly do add a lot. Back in the days of Counterstrike, the better players memorized the sight lines and hiding places. With destructible environments, those hiding places are but a respite that may not exist 30 seconds later. It certainly adds a sense of urgency to the game.
I, too, welcome new innovation. Pretty graphics are impressive, but they make no difference to me in-game (other than often getting in the way). Bloom, smoke, camera shake: sure, it may be more "immersive", but it also gets between my sights and the thing I am supposed to be shooting. I'd rather have something that adds to the fun, like game modes, levels, new weapons and perks, or something entirely different than nobody has thought of yet. New and different is part of what makes a game worth playing. Fancy graphics just make for impressive trailers.
Old way + more polys: low risk, content pipeline is in place.
New way: tons of risks, whole new kinds of bugs to explore, possibly a whole new content pipelines to be built.
You're the guy with a big bag of money. Which would you invest in?
People focus on realism, but i think that while realism is a very good thing to achieve, the games can do more than that. As long as a game is believable, it can get away with non-realism.
Beyond that, i avoid dynamic/open-ended games (i'm talking from a gameplay perspective - blowing stuff to pieces and having good interactivity with the environment, especially beyond just kicking barrels and boxes around, is something i like) because most of the time i feel lost. I prefer more linear games, although not to the point of telling me where to go. Probably a mix between linear and exploration (ie. the game is linear, but you have to find your way yourself to the next point). Obviously this needs good level design that helps the player and isn't against him (like not being a victim of Area Copypastiosis).
Talking about good levels, how about spending a little of those "bling" money to build more levels for games that aren't finished in 4-5 hours?
come on! how come raising water level in a tank should be any problem at all? f.e.a.r. had no trouble with it at all. and also buildings were completely destroyable in Mercenaries 2: world in flames(they fell to pieces as they were collapsing almost 9/11 style).
I know XCOM Apocalypse was the mutant child chained in the basement as far as most fans were concerned, but the fully destructible environments really added to that game. And it made grenades fun as hell.
Aliens were hiding in a conference room waiting to ambush you, so you had one guy blow a hole in the wall and your other guy toss some grenades in while avoiding the line of fire.
You could do that in the original XCOM, too. Aliens just don't seem to understand how poorly the walls of country farms hold up against guys with assault cannons.
Good (or at least interesting) AI is the most exciting feature a game can have, as far as I'm concerned. Bethesda's Radiant AI isn't always on the ball, but it usually fails in a way that's more "crazy" than "stupid". I love it when the computer does something I'm not expecting.
The problem is that most developers want their games to look as pretty as the machine can handle, because otherwise their game will look dated. That's the issue. I'm pretty sure the PS3 would be able to handle that train level without problem if the graphics were a lot worse. But then the publishers would throw a fit. Unfortunately, that will never end until the current gen graphics actually become so demanding that it's just not worth it to push them to their full effect, which I predict will happen relatively soon.