I once had an interviewer with two (TWO!) PhDs. He made sure that I knew he had two (TWO!) PhDs by handing me his business card as we sat down and casually remarking that he had two PhDs (see, right there, two of 'em, yupperoo). This behavior ("he's kind of jerk, and he has two PhDs") had been accurately foretold by the prior interviewer (that session had gone great).

The guy-with-two-degrees asked, "What is the simplest way to synchronize two threads?"

Well, "simple" is pretty fluffy and subjective, so I asked what he meant by it.

"You know, the most simple way."

It didn't get better. I probably made a mistake, and started naming a bunch of synchronization schemes. "Mutex? Semaphore? Dekker's algorithm? Spinlock?". Each mention got a response like, "No, I mean simple. Simpler!"

I never got it. The rest of his questions were similar ("What is the best way to do RPC?"). His parting words to me were, "You need to go back to school."

The specific answer he was looking for was, "mask off interrupts". Doesn't work on multiprocessor systems, so I'd not even mentioned it. I wrote their hiring manager a polite email along the lines of "thanks for the interview, here's one question that I got wrong and why."

No surprise, I didn't get an offer.

Four or five months later the firm called me back, saying that they had fired the jerk and was I interested in interviewing again? I told them I was pretty happy where I'd landed. (I did not congratulate them on firing the jerk, and I suspect they'd had trouble hiring anyone while he was there).