The major problem I see with this is that while the computer itself might still boot, the world around it will have changed so radically it will effectively be useless. You can see this with machines much younger than 50 years, they can't support the encryption required by most modern websites, interfaces age out of regular use: used a 24 pin parallel port lately? How about a modem? Even software with long support fo…
Keeping them relevant is more of a challenge, this is one reason I hate to see people forcing redirects to HTTPS, it's just great and dandy that modern browsers suggest this to users (as long as they don't force), but for vast vast amounts of information, in the free world, you shouldn't force people into always-encrypted. There is zero good reason I shouldn't be able to choose to read Wikipedia unencrypted. And I can do that on pretty much any machine that can somehow crawl onto the Internet.
There's also the problem of "easily replaceable parts", because parts stop being produced, and I think we've been in a golden-age of new-old-stock, and I don't see that continuing in the future. Easily replaceable pose two problems, first, is sourcing the actual part as the world moves on and production stops, the other is that easily replaceable is at odds with industry demand, sure, I personally like 40 pin dip sockets for my hobby electronics, because they're easy to solder, things are large enough I can see them! But I'd not want every appliance I own to be constrained by that, meaning there's no incentive to produce them, leading back to the first problem.
There are so many problems with how to build a machine that lasts, or, even how to build an architecture, or system, that lasts, but they are worthwhile pursuits, discussing and thinking about them is worthwhile, because, systems we can understand, own and service are inherently more valuable than the ones we cannot.