The only thing that becomes clearer with each data point is the true randomness of this process. I have heard different things from different schools about their review process, but I am sure of one thing, which is that no admissions committee (which is what, 10-15 people?) is reading all of the several thousand applications they receive pre-interview. All our applications are probably divided, packaged, and distributed to screeners throughout the university or hospital system, and I wouldn't be surprised if the way interviews are distributed is based on the 'scores' or 'recommendation' (i.e. recommend to interview, no interview, or hold) from the screeners you happen to be assigned to.
This screener system is probably what introduces so much variety into the process because different people have different standards. My n=1 data during a few interviews showed me this. At one interview, my faculty was raving about how volunteering I had done, more than she had seen for the other people she had seen this app cycle (she was a very new faculty and told me this was only her second year interviewing). At another interview, I was paired with a retired faculty who was once on the admissions committee, and one of his first questions was, "why did you choose to pursue less time volunteering during college?" So, to him, the same amount of volunteer work that had impressed the first faculty was minimal. The same amount of variability is likely rampant in the screening process! Some of us probably get assigned to screeners who vibes with our application a lot at one school, and happen to be assigned to others with different standards or preferences at other schools. And this is also where the human bias element comes in huge. If I got assigned to a screener at Stanford who had a bad experience with colleagues from "insert X subspeciality," they might be less inclined to vouch for an applicant who has modeled their app around the same subspeciality. And vice versa. And then, of course, all the social issues that play a role in human bias. For example, I wrote much about my family's religious experiences for diversity prompts. At one interview, my faculty and I went a half hour over time talking about religion in medicine in detail and our shared experiences (they were from a different faith, but you can still find commonality). In another interview, my student grilled me on whether my religious beliefs will affect my medical practice (totally valid, but clearly, they had different feelings about me choosing to share that part of myself than the former faculty). And lastly, even choices such as whether to include a hobby on AMCAS can have an impact. I talked to two admissions reps in the Spring from two different schools, and one of my questions was, "should I include this hobby?" One gave me a resounding yes, another a pretty hard no. Guess which one I have an II to right now?
That's the only way I can make sense of why some people will get an R from Case or something and end up matriculating at Yale. Crazy stuff.
Post-interview when the actual committee starts reviewing and making decisions is where I think the process becomes much more predictable.
I would be in the wrong if I did not end by saying this is an outlandish amount of speculation based on the "grapevine." I could be very mistaken.