I think we must decide if it is a preprint server or publication. I get the aim to preserve signal to noise ratio and facilitate discovery, but think this will naturally emerge as the noise floor increases. Either via credentialism/names/institutions we filter by, or increasing depth in niche areas.
