Like the parent comment by derefr, my actual, non-hypothetical practice is saving to the Wayback Machine. Right now I'm probably saving things at a rate of half a dozen a day. For those who are paranoid and/or need offline availability, there's Zotero https://www.zotero.org. Zotero uses Gildas's SingleFile for taking snapshots of web pages, not PDF. As it turns out, Zotero is pretty useful for stowing and tracking any PDFs that you need to file away, too, for documents that are originally produced in that format. But there's no need to (clumsily) shoehorn webpages into that paradigm.

If you do the print-to-PDF workflow outlined earlier in the thread, you'll realize it doesn't scale well, requiring too much manual intervention and discipline (including taking care to make sure it's filed correctly; hopefully you remember the ad hoc system you thought up last time you saved something), that it's destructive, and that it ultimately gives you an opaque blob. SingleFile-powered Zotero mostly solves all of this, and it does it in a way that's accessible in one or two clicks, depending on your setup. If you ever actually need a PDF, you can of course go back to your saved copy and produce a PDF on-demand, but it doesn't follow that you should archive the original source material in that format.

My only reservation is that there is no inverse to the SingleFile mangling function, AFAIK. For archival reasons, it would be nice to be able to perfectly reconstruct the original, pre-mangled resources, perhaps by storing some metadata in the file that details the exact transformations that are applied.

Archives are only worth their weight in water if interested parties can find what they're looking for. When artifacts aren't gathered and curated into public-facing unities or collections, then history elides them until further notice. These threads are still floating in the sprawl of the Twitterverse, placed into history and drowned out by an ocean of pure, frantic noise. What this piece makes evident to me is the need for restoration: that they need to be resurfaced, preserved, made visible again.

Given that this document cites a number of non-persistent web resources, I have archived a copy of https://wellcomeopenresearch.org/articles/4-170/v1 at http://web.archive.org/web/20191224000829/https://wellcomeopenresearch.org/articles/4-170/v1 using the "Save outlinks" mode.

Probably a good idea to do this routinely for all articles in the journal.

Archiving service with an emphasis on scholarly publishing.

Archiving pages that block it.