Same reason they're used in grading. How else do you ensure that "partially correct" is consistently weighted?

Certainly, in college, I recall one calculus problem, we had to find the local maxima or minima of a function, something along those lines. Long and short, I differentiated, plugged in for zero, and made an algebra mistake, got one right, one wrong. Something like that. 5/10. Someone else...differentiated, and ended with the differentiated function set equal to zero, but didn't solve for it. 8/10.

No idea if that was a lack of rubric, or a badly worded rubric ("got one of the maxima/minima wrong = -5 points", "Well, he didn't get either one wrong, so..."), but yeah, a good rubric would have prevented that inconsistency in grading that I still carry with me to this day because it was so unfair.

No, I don't. Because engineering mindset isn't something you can quantify, and I don't care if a senior position isn't able to rattle off all of the 2xx HTTP status codes from memory. Having a conversation with candidates instead of ticking off boxes gives me the absolute strongest signal every time.