I don’t know, real-time graphics programming is and has always been about hacks. Today it’s perhaps not often hardware-level hacks, but the ethos is still “cheat as much as you can, and then some more for good measure”.
Which of today’s real-time demos or games are you thinking of? The trend over time toward physically based rendering over the last 40 years has been pretty clear and absolute, and your claim is failing to capture that fact. While it’s a little sus to frame all real time graphics programmers has having the same goals, whatever ‘ethos’ does exist in common across all graphics programmers, history demonstrates that it includes the goal to increase physical realism, which I would call reducing the hacking/cheating. Wouldn’t you? Today we have multiple AAA games on the market with path traced global illumination and physically based materials.
Not really - graphics cards are now powerful enough to ray-trace every pixel every frame, if you are doing it with a reasonable SDF instead of a huge bag of triangles.
Meh, real-time reflectance functions are still going to be all sorts of approximations. To say nothing about stuff like real global illumination. And SDFs are really nifty but a function that represents a complex, detailed, non-stylized/cartoony scene such as those in modern games is not going to be fast to evaluate.
Not to mention other aspects of the overall visual experience, eg, everything about scene dynamics, object interactions, etc. A bigger compute budget is always welcome.