As the paper points out that's accounted for in the analysis, because shockingly enough to most internet commentators researchers know that they need to control for confounding variables
I don't think many realize how huge the replication crisis is. You have 75% of studies in a leading social psychology journal failing to replicate. [1] Another replication effort on preclinical medical studies found only 11% were able to be replicated, a replication effort on cancer studies found the average effect size to be 85% lower than published, and much more. If you assume most science headlines, at least outside of the traditional hard sciences, are false - you're substantially more likely to be right than wrong.
[1] - https://en.wikipedia.org/wiki/Replication_crisis#In_psycholo...
This is very true. In my PhD research, I managed to perfectly predict the placebo effect using a random forest. Now, this was actually due to sampling variance, and the finding went away when I averaged over splits.
I often wonder if I'd still been in academia when this happened, if I would have been able to resist the siren call of a Nature/Science paper (and the consequent career benefits).
I doubt OP's jaded response was alluding to your insight, but still thanks for that, it's genuinely good to know.
'Why ‟controlling for a variable” doesn't (usually) work' - https://dynomight.net/control/
Other things that don't work, including four other 'controlling for a variable' entries at items 39 through 43: https://dynomight.net/things/
Dan Olweus did several of them in the 70s and 80s, and fought to his dying day against programs which didn't have casual intervention studies backing them.