The point you're raising is at the moment important, but ultimately irrelevant.
The scientific paper in digital format when it was first produced, did not have all the advantages of physical paper. Physical paper was easier to archive (digital storage would corrupt easily), easier to move around (you needed a computer to view a digital paper, and most people didn't have one; networks were mostly non-existent), and easier to annotate.
These were all important things to consider at the time, but the technology eventually caught up, and surpassed printed paper in most of not all aspects.
Keeping dynamic content consumable through the years, would be costlier than keeping static content consumable. But the price isn't that high. Web standards are designed with backward compatibility in mind, and the software you use to view web content (browsers) is mostly open source. I'd imagine it'd be much easier to view web content produced today in the 2030s, than it is for us to play NES games produced in the 80s.
Not all paper (papyrus, wood or stone) based documents remains till this day. Effort was made to conserve it because, through different times, enough people thought that knownledge was worth being remembered.
If a interactive paper is meaningful to enough people, it will be preserved.
I think you're conflating two distinct issues. One is whether the media is static or interactive, and the other is whether it is a page on the web or stored in some other fashion.
If the scientific papers will be around long-term, that's likely because they'll be stored in some sort of database. There's no reason why interactive media couldn't be similarly stored.
(Yes, there are issues about being able to run the interactive content in the future, though you don't bring that up).
I’m not so sure about that. Web standards have made it so that the browser you install today can still view the websites that were made when your favorite Super Nintendo game came out.
And you wouldn’t print just one copy of a paper ever, so a network of database servers getting destroyed is the equivalent of a bunch of physical libraries getting destroyed.
Flash may have gone away but Flash was not a core component of the web spec, like HTML. It was a third-party plug-in.
In a world where the PDFs are being backed up, the videos are also usually backed up and are just as available. I think in most cases, having an animation gets most of the value that you'd get from having complete interactivity.
If it disappears it will be because nobody kept it up. Not because your web browser is incapable of rendering them in a decade or two.
An interactive webpage can be self-contained. It doesn’t have to load cross-origin resources, it can simply include them in the directory. And if it’s a scientific paper it should have no need to load data from external sources. An interactive scientific paper can absolutely be self-contained.
Anyways back to the main topic, if it disappears in a decade or two it will be because nobody kept it up. For example if it was hosted on some random guy’s domain who eventually didn’t feel like hosting it anymore, etc. But even physical paper-format scientific papers aren’t just stored in random people’s garages, they are also archived by other organizations. So if archives like the library of congress kept their own copies, then you would expect the content to last as long as the library of congress (or other archives) would want them to.
For similar reasons, given enough time I'd expect results from a paper to be in text books and common knowledge, or be forgotten. Either way nobody would learn it from the paper after a few decades.
You can't be a researcher in a field and only read articles from the last 20 years. I've lost count of the number of times I've been struggling with a problem, come up with a partial solution, then found a nice, complete, well-thought-out solution in some random engineering journal from the 40's. Use the literature. Never, ever, ignore papers just because they're from a few years back. You may need to filter the ideas through a few paradigm shifts, but the underlying observations and methods can be very useful.
Textbooks can be wonderful overviews and great ways to learn, but they only scratch the surface of topics. Furthermore, in most cases, you can't directly cite textbooks. Only original research is citeable. Review papers, most textbooks, and other compendiums are not citeable in any field I've been involved with. The exceptions are textbooks that have original research in them, which are very common, but are more akin to special editions of a journal than most textbooks.
This often leads to researchers reinventing the wheel. Good ideas can fall out of favor essentially randomly for indeterminate amounts of time, so there should be no expectation that what is cited today is necessarily the best.
To use another example I've seen a few times, an approach that might have seemed like a dead end in the 50s could now be tractable due to other developments, like modern computers.
I'm going to cite Codd 1970 but I learned SQL from web tutorials written 30 years after the paper, which I've only skimmed.
But every scientific paper ever published will be around until the end of humanity, whereas this interactive web page will surely disappear within a decade or two. The scientific paper - a self-contained entity that can be moved around, printed out, archived, searched and annotated - is an incredibly powerful entity.