As a fun warm up-question, I always ask the candidate what editor and keyboard they use. The candidate who geeks out about their plugins and cherry switches inevitably passes all 5 rounds; the candidate who uses eclipse with maybe a vim plugin will go on to fail about 75% of the time. It's a good filter too; I'd say 1 out of every 20 candidates expresses strong convictions.

I'm seriously considering using it as a dog whistle and skipping the experience deep-dive (which tests if you are a good story teller), the systems design questions (which tests if you can sync to the interviewer's sense of judgement), and the algorithm hazing (which tests social anxiety). But I'm worried there is some kind of discriminatory correlation baked into that metric.