There's earlier papers (in PNAS?) showing the same thing. One of them at least demonstrated that the correlation between impact and dropout was fairly low, suggesting that scientist dropout was not very closely related to quality of contributions or something.

I have no doubt in the paper's conclusions but think science is underestimating the implications, in terms of attributions of credit or reasons for success.

Can you remember the title? Would love to have a look at the analysis.

I think it was this paper?

https://doi.org/10.1073/pnas.1800478115