I mostly agree about catching fraud*, but catching mistakes? I didn’t know that it was a mainstream view that that’s beyond the purview of peer review. Of course it won’t catch them all, but catching mistakes should be an aim of the peer review process.

* though I do think journals should take steps towards basic integrity checks, eg, computational reproducibility, but do it in house, not relying on reviewers. While those won’t catch most fraud, the errors detected could lead to extra scrutiny, some of which might uncover (or at least deter?) fraud