It is radical but my idea is that one submits to ranking houses (which could remain the existing journals). Every paper submitted will be published online and ranked. You may submit a paper only once. You fix it up in light of (presumably more careful and more numerous) referee reports, but then it is published with a numerical ranking. Advantages include 1) less refereeing overall and so, potentially, more careful refereeing from people who more closely specialize in the area. 2) much quicker time from submission to publication, eliminating pressure to choose where to submit on strategic grounds, 3) encourages people to finish papers before submitting them, rather than treating submissions as entering a lottery 4) as is the difference between just in and just out of a journal is enormous–this system allows one to get credit for “very close to making it into Phil Rev”. Call this the Wine Spectator Model.
However, a serious concern about this proposal is that one bad set of reports has a more dramatic effect on a paper on this scheme than under the status quo. As is, one can get a set of terrible referee reports and recover to still have the paper placed in a good journal. That, I acknowledge is a real issue with this proposal.
But I think people are not appreciating the scope of the problems that currently exist that need fixing. Many humans are spending a lot of time, for example, desk rejecting a lot of papers and all that would be eliminated. Many papers take years to come out and people are agonizing over whether they can afford to send the paper they love to a top journal, given the high rejection rates and the need for pubs before hitting the market. The strains on referees and those seeking referees is too high.
Some may also think an additional benefit is that it would be good for the profession if we produced less but better papers. This would certainly make reading the literature and refereeing more rewarding.