I think the idea is that a candidate might have a bad interview or not click with a given interviewer, and so doing a few interviews averages this effect out a

I think the idea is that a candidate might have a bad interview or not click with a given interviewer, and so doing a few interviews averages this effect out a bit. At least at Google, every interviewer has their history and overall calibration tracked, where the calibration is, I believe, something to do with how much your scores match the other interviewers' scores. This is probably fairer overall, albeit more expensive for the company and the candidate.