I agree with your answers, but I would like to point out that a few dollars per year isn't necessarily a trivial amount depending on who you ask. And top-level domain fees are more or less flat across countries, besides a few exceptions.

I had never heard about geocities, that's interesting, thank you for sharing this.

Now, to elaborate a bit more about what I said, I don't think there is a lack of technical solutions for this, but rather a fully integrated solution is missing. I would like to be able to "seed"/mirror specific content, like articles that I find interesting. Of course, that wouldn't help with content I (or others) haven't read already, but you would be responsible for the content you host, as you pick it (unlike ipfs or freenet, AFAIK). Integrated in a web browser, I suppose it could also "mirror by default", though I would be concerned about leaking user-specific content (so maybe a spec/protocol extension would be necessary for html/http to tell what could be mirrored?).

It occured to me that this was a bit like "boosting" a post in mastodon. Indeed ActivityPub could be an interesting transport mechanism for this, and could be a way to propagate update notifications back. Though I am not sure the content itself is mirrored with the current mastodon "boost" implementation. A mirror index could also be ran over DHT, while a "search engine" could provide the post hashes, and let clients fetch the content from the DHT.

The whole idea is still quite a bit rough, but it surprises me that in 2019, hosting plain text content (more or less a few images) in a future-proof way isn't a solved problem, and we still have to rely on the Web archive. Applying trusted timestamping (by a few trusted third-parties) to the web archive content would be a good start, to trust mirrored content.