If you can't calculate the optimal design (for hiring, in this case) from first principles, what option do you have but empirical observation? And when steady-state performance takes at least 2 years to obtain, is it unreasonable to have fads at roughly that frequency?

Resumes suck, take-homes suck, interviews suck, nepotism sucks; yet people still need to invest $x00,000 based on something. I don't have the answer, but let's not pretend it's not a hard question.