A journal or conference that publishes everything is useless to its audience, which expects significant curation and filtering to pare down the incoming firehose to something that (a) fits a specific topic and (b) has a filter for quality and (c) pick the most interesting ones out of all the far-too-many decent papers that fit the first two criteria.
There's no problem to publish them all somewhere, that's being done on e.g. arXiv, pretty much every recent useful CS paper is there as well, often before being published in a "proper venue" - but a "publish everything" service is not enough, people do want to read more selective sources instead of trying to skim everything that's posted on arXiv.
These days nothing is keeping people from publishing elsewhere if they want to.
There has been quite enough censorship and paywalling of research that my taxes fund.
Since a good chunk of researchers are funded in part by the government, I guess most journals would end up having to follow this "publish everything" requirement (money is fungible and some of every grant goes to administrative overhead, so you could argue anything any most universities touch is funded in part by the government).
Most publishers publish multiple journals, though, so I guess they could follow your rule as long as they were allowed to open up the, for example, "IEEE Journal of Perpetual Motion Machines And Straightforward Proofs That P=NP."
Why not publish them all? Endorse those that are selected, add commentary to those with which there is disagreement, but is a batch inclusion of them all so technically difficult?