No, they can’t. The “caching inefficiencies” (?!) are part of the hardware and can’t be optimized away by a JIT. I think you have a very confused understanding of the post.

A compiler (JIT or AOT) can certainly make decisions which are more or less cache-friendly. It does depend on the language and the amount of freedom the compiler has, but just as one example, struct/object field reordering can reduce padding and thus fit more useful data in a cache line.