Would it not be possible to solve the problem of "thread migration to an incompatible core" on the kernel scheduler level?