I am not sure if raytracing has a future. So far the comparisions I've seen are all very underwhelming, sometimes I can't even see a difference. Which is not good for a feature that costs so much performance. I've heard that the next console generation comes with raytracing support so I could be wrong there.
The same could have been said about the first hardware accelerated 3D games. The difference between Descent or MechWarrior 2 with 3D acceleration was a texture layer and occasionally a few more polygons, both of which were done in other incarnations of similar games without any 3D acceleration at all. The last time I played a 3D game that even included a software renderer was probably back in the late 90s. Even word processors are 3D accelerated now.
It's simply a matter of performance. The effect underwhelming now, because RT performance is still underwhelming. There has to be provisions for a rasterized fallback, so nothing can really depend on the raytraced elements, which must be used sparingly anyway, to maintain acceptable performance. This will not always be the case.
Dedicated RT hardware probably isn't much of a solution in today's market, but better ways to use more general purpose shaders for the purpose of ray/path tracing most certainly does.
Anyway, if you've ever paid attention to ED's shadows, it would be clear that there was enormous room for improvement, and that even a basic RT implementation could do wonders (though this is hardly the only way they could be improved). Of course, I still don't expect it to happen for this game.
I wonder if NVIDIA have actually jumped the gun on this one? - if raytracing in underwhelming now then consumers may get the idea that it isn't much of an advantage and it'll die the death of "no demand".
They had no real choice with the timing.
They found themselves in the rare position where they had such an advantage over the competition (AMD currently, but soon to include AMD and Intel) that they could afford to blow a third of the transistor budget of some of their parts on hardware that would be essentially unused in traditional titles, while still retaining performance leadership in those titles.
If they had waited and found themselves in a more competitive high-end market, devoting that many transitions to RT would have be disastrous.
Even if their dedicated hadware approach doesn't go mainstream, it's already been a useful marketing too, and has already influenced the rest of the graphics ecosystem.