Also, a heavily pre-selected population (as evidenced by the excellent OS in the non-responder arm) who had a complete surgical resection prior to starting treatment.

Still very exciting, but not as gobsmacking as the RFS KM curve appears.

I don't think that's fair. The preprint has methodological limitations but is quite similar to preclinical studies on Long COVID, which were also often shared and promoted prior to peer review. I understand your concern that the findings will be weaponized, but censoring science is a slippery slope.