> Evolving a new spike is extremely disadvantageous for the virus.

... Right up until almost everyone has an immune response to the spike and not to any other feature, at which point a virus which changes the spike and potentially keeps its other features is a winner. We are creating an immune monoculture.

> ... which hinges completely on the virus not evolving at all, which is outside of the realm of reality.

This makes no sense.

If you only only target the spike, you assume the spike doesn't change sufficiently. If you target the entire virus, you assume the entire virus doesn't change sufficiently. As far as I know, mutations are generally assumed to be random. You're effectively matching against a shorter string (spike only), so you're less likely to recognize a mutated virus, than if you match against a longer string (the entire virus) which is what natural immunity targets.

Mutations in the spike (that still keep it effective - and we actually have quite a few of these already) are much more likely to evade the vaccine-induced immunity than they are likely to evade natural immunity.

Mutations only in the nucleocapsid are more likely to evade natural immunity than vaccine-induced immunity, but to a lesser degree - because natural immunity also targets the spike.