Ultimately this is a problem that's never going away until we replace URLs. The HTTP approach to find documents by URL, i.e. server/path, is fundamentally brittle. Doesn't matter how careful you are, doesn't matter how much best practice you follow, that URL is going to be dead in a few years. The problem is made worse by DNS, which in turn makes URLs expensive and expire.
There are approaches like IPFS, which uses content-based addressing (i.e. fancy file hashes), but that's note enough either, as it provide no good way to update a resource.
The best™ solution would be some kind of global blockchain thing that keeps record of what people publish, giving each document a unique id, hash, and some way to update that resource in a non-destructive way (i.e. the version history is preserved). Hosting itself would still need to be done by other parties, but a global log file that lists out all the stuff humans have published would make it much easier and reliable to mirror it.
The end result should be "Internet as globally distributed immutable data structure".
Bit frustrating that this whole problem isn't getting the attention it deserves.