I don't really think it's fair to compare game engines with the demoscene. The demoscene is really about exploiting hardware to the fullest extent and that's not really an economic advantage when you get new GPUs every year that are 50% to 30% faster. now, creative hacks can increase performance several hundred percent, but is this sustainable for a product? probably not. certain hacks are very specific and that reduces your market. they’re usually non-portable, fragile, or locked to very specific architectures. That doesn’t scale across consoles, PC variants, or mobile. So in a market context, the value of raw ingenuity is absorbed into engine pipelines or middleware, not showcased in standalone demos.

Now that is to say a demo can't be portable, for instance I've run several older demos on my computer. I've also had instances where demos wouldn't run on my computer. Imagine if that was a game that I bought. Or imagine if it was anstracted to such a point that the feature is just not available dynamically. For instance of particle effect became 100 times less impressive.

yeah, companies wants little effort for bigger effect. They want maintainable code, etc. Clever hacks are not really welcome.

Now that PC's are so powerful, you look for algorithms optimisation. And although it requires a ton of cleverness as well, it's much more documented than obscure hacks. You learn that at school.