> there is nothing in the universe that says "there should be capsaicin".

There's nothing in the Universe that says "there should be capsaicin" specifically. But conditional on animals having particular receptors, there is something that says "plants with it can outcompete plants that don't".

Yes, evolution has (significant) random elements in terms of where you start on the fitness landscape and in terms of non-biological factors (e.g. "oh shit a meteor just hit the Earth) that can sometimes intervene. But evolution is tightly intertwined with game theory, and it isn't a coincidence that game-theoretic strategies show up all the time in evolutionary biology.

"Organisms, broadly speaking, will eat and reproduce" is just as iron-clad a law of our Universe as "objects will roll downhill" is. Maybe more so, since it's implied by abstract mathematical law and not even by the particular quirks of actual physics.

> There are, for all practical purposes, infinitely more traits that have never and will never be expressed than ones that we've ever seen.

This is true to some extent, but the frequency of convergent evolution shows us that some patterns really are just super useful. Wings have evolved independently in birds, insects, mammals, and even plants if you count the little fins on a maple leaf. That's not a coincidence.