https://en.wikipedia.org/wiki/Webring s used to be quite useful, maybe along with https://en.wikipedia.org/wiki/Planet_(software)
Other than that, it's a good question. It's been explored by many that Google tends to forget old content. Alternative search engines, such as DuckDuckGo, in my experience, tend to lean towards indie sites more.
EDIT and, of course, there was https://en.wikipedia.org/wiki/Technorati
A federated curated directory, perhaps with an emergent shared vocabulary for tags/categories/etc. (similar to the Wikipedia-centered naming guidelines for wiki pages) might work. The DMOZ model wouldn't scale in the modern web, with its rampant contentiousness around politics etc., and even the original DMOZ broke down because of corruption, with websites having to pay category "editors" for listings.
There was even a tip level directory page of unusual devices connected to the internet. Like there was one single fishtank, one coffee maker and one internet connected vending machine... and everyone thought these things were marvelous and delightful.
I know it would get gamed really, but what if there were an injected "related-sites" link list as an extension, that drew from a curated directory? That was how you used to explore, through blog and site "friends of" sections that sent you on an endless rabbit hole.
I feel like it can't be feed oriented, because that's a more consumption mindset than exploration mindset, and I feel part of the fun of the old web was that constant active exploration.
I don't want to turn this into a conversation about SEO, but suffice to say commerce dominates search results for so many terms that it can be hard to find anything community oriented, or niche information. So search engines and feed sites are not quite right for exploration.