I'm not trying to defend it (and as a hiring manager I don't do it) - I'm just explaining the implication of Chesterton's Fence as applied to interviews. "Processes and tests should serve obvious purposes" seems very wrong to me, though - they should certainly serve useful purposes, and those purposes should be documented, but there is no real reason they need to be obvious to a layperson. Google specifically dropped those riddles because they did not find them effective, which means they have some metric for testing efficacy. They have not dropped algorithmic challenges, though. I wonder why?