I read the new paper and I'm familiar reading physics articles. It strikes me almost as GPT-3 generated in places.. It certainly would not be accepted in its current form by any peer review process in a high impact journal. I'm sure the intent and ambition is absolutely sincere and there is serious research here but it's presented in a very unclear way IMHO. I think the paper could be rewritten to be more focused..
What the heck are peer review scores? And why would you expect it to be correlated with citation count? High quality research isn't the same thing as highly citable research. Peer review is supposed to check plausibility and maintain minimum quality standards. Peer review isn't a sign that something is good. But if something can not even get peer review it's a sure sign that it's pretty bad.
I know that video. This is also different from pure peer review because it involves an editorial decision/recommendation. I am sure >95% of the papers rejected from NeuralIPS will pass peer review elsewhere.
So this isn't about getting published in peer reviewed venues, this is about getting published in a particular, selective, high visibility venue.
So as far as I can see none of this has any bearing on anything you replied to.