Comment by duskwuff

> "Random sample of Internet links" is going to include a lot of absolute garbage.

It's also likely to include a lot of non-content links, e.g. links to index and navigation pages, interstitials, search results, user profiles, image galleries, etc. These sorts of links don't reliably address specific content, and it's natural that they'll change or die over time. This doesn't necessarily mean that anything valuable has been lost.