I think this also applies at all levels. I'd argue that it is the big reason people feel very frustrated with reviewing, and especially in hyped areas (e.g. ML). There's plenty of incentives to reject papers (low acceptance rates mean "higher quality" publications, advantage yourself by rejecting or being overly critical of your competitors, you get no clout for reviewing and no one will get upset if you take all of 5 minutes to review), but very few incentives (I can't even name one) to accept papers. It is fairly easy to dismiss papers as not novel because we all build off the shoulders of giants and things are substantially more obvious post hoc. Metareviewers and Area Chairs will 99/100 times stand with reviewers even if they are in the wrong and can be proven so (I had a reviewer give me a strong reject claiming I should compare to another work, which was actually our main comparitor and we compared to in 5+ tables and 5+ graphs). I can't see these issues being resolved until we all agree that there needs to be some incentive to write high quality reviews. The worst of this is that the people it hurts the most is the grad students and junior researchers (prevents graduating and career advancement). I'm not saying we have to accept papers, but I am saying we need to ensure that we are providing high quality reviews. Rejections suck, but rejections that don't provide valuable feedback are worse.
If the publication system is a noisy signal then we need to fix it AND recognize it as such. There's been plenty of studies to show that this process is highly noisy but we're all acting like publications are all that matters.
This is all before we even talk about advantages linked to connections even in double blind reviews, collusion rings, or citation hacking. I feel we can't even get the first step right.