[Someone]

Authors are to blame, too. They can put their publications on arXiv. https://www.elsevier.com/about/policies/sharing:

“Authors can share their preprint anywhere at any time.

If accepted for publication, we encourage authors to link from the preprint to their formal publication via its Digital Object Identifier (DOI). Millions of researchers have access to the formal publications on ScienceDirect, and so links will help your users to find, access, cite, and use the best available version.

Authors can update their preprints on arXiv or RePEc with their accepted manuscript.”

there still is an issue with older articles that authors don’t have PDFs of, or whose authors aren’t in the field anymore (or even died). That problem will stay, but _if_ authors start putting their publications on arXiv, their university server, etc. en masse, also will get smaller over time.

[stared]

Leading journals allow publishing preprints on arXiv only as there was considerable pressure. In the quantum physics community (where it was expected to post results on arXiv), a journal not allowing arXiv was not considered. So either their de facto allowed, or even started officially to do so.

I don't remember the exact dates, but Nature (who had been hesitant for a long time) gave in as well. They resisted a long time as they were (and are) considered a badge of honor.