Another data point - I'm an EM. I still get all the usual SWE interview questions (system design, coding which is often leetcode, behavioral/cultural, etc), plus additional loaded ones about how I've handled various situations in the past (which is always fun because sometimes they're about things I've never had to handle because the question assumes something that shouldn't be true, i.e., "describe a time you had to handle an employee who was angry with you", well, that's never happened; well, why is the employee angry with me, specifically, what happened to all the time spent building trust so they could calmly come to me with issues that they had, all the times they saw me advocating for them so they know that a disappointing and unexpected outcome is due to factors I'm not in control of, why had I dropped the ball in being responsive to their earlier voiced concerns, etc etc). This latter usually is asked by the hiring manager (usually a director)
My last three jobs, at least, I landed because I clicked well with my immediate hiring manager in this particular interview. That was it. I did no worse, and no better, than I did elsewhere on anything maybe a bit more objective (system design is weird, but coding, at least, you can say "does it work/is it efficient or not").
Which sounds bad, but it means my expectation of management aligned with theirs. It was a complete crapshoot in every case, with a lotta rejection along the way, before I found that, but each time it led to good outcomes.
Interestingly, there was also one company where I know I was a strong contender, did decently with the HM, but was beaten out by another candidate. I know because the recruiter reached out to tell me how close it was, and that was further validated when I later applied to a role, spoke with a different recruiter, who said "yeah, I saw the feedback and was like 'why the heck didn't we hire this person?!'". But that second application went nowhere; the HM gave the initial phone interview and didn't want to continue.
It's all random (or at least, so many variables that can't be measured or controlled for as to be indistinguishable)