I associate that with bad-faith attempts to prove a point, which I don't think this is. It feels more like a communications problem where a finding like 'Most

I associate that with bad-faith attempts to prove a point, which I don't think this is. It feels more like a communications problem where a finding like 'Most (65%) cases respond favorably to X', followed by lots of hype around X, only to see a bunch of surprised pikachu faces when a subsequent paper looks at the 45%. Maybe it's just hype cycles, but it feels like there's gotta be some professional obligation to proactively consider the outliers/edge cases/minority group(s), rather than waiting several years for the "woah wait, turns out not all X are Y" paper.