[mnl]

That's very good news, yet I imagine the publisher can claim rights over composition, layout, typesetting and the pdfs themselves. This is quite a problem as who's going to do all that again, particularly when it comes to dead authors (ocr won't cut it). Nowadays you have your own digital version and you could simply distribute that (if we ignore keeping track of published page numbers), but what you get for papers up to more or less the 90s is scanned documents, it's unlikely that you can find other versions around. The transition to a sane model won't be frictionless nor fast, it might be cheaper to make a bid for the publishing companies outright (fees of subscriptions are simply absurd, their whole business model is a legal racket).

[zaarn]

For dead authors, the inheritance regulates who gets to control the copyright. Usually inheritance works out, the state puts a lot of money into finding lost relatives if everything else fails.

The publisher to my knowledge cannot claim copyright over composition, layout, typesetting or the PDF unless they can show they had significant copyrightable work in each step. (Copyright in Germany first requires you to do some significant intellectual work, where significant changes based on what you do but science usually has a higher barrier)

If the PDF contains a trademark of theirs then that's a problem.

Generally a pre-print PDF should be available in almost all cases and could be published without Elsevier's input. If not that, Universities do archive the Tex files (or similar inputs) if possible so they could take those.