Then it's circular: "only the units that are publishable are published"

If it's not worth publishing, it's not a result at all. That being said, the problem really is that since LLMs excel at sifting through piles or papers, publishing more would aggravate the problem instead of alleviating it.