No real reason except no one has tried yet. It's extremely difficult to get search right, although Moore's law and tensorflow has probably made it easier to attempt today. Latency would be significantly worse than major search engines though. I think for the people who care about the open web, having a libre alternative search engine should be the highest priority. It's one of the few classes of software/services that no one has tried replicating open source.

Hmm, doesn't eMule have something like that in place (and nearly 2 decades now) ? (It's more janky than a centralized search engine of course...)

No, you don't need a search engine for that. Blogs and YouTube channels commonly link to similar channels and content. A search engine is a form of content discovery, but it's not the only one, and technically we already have a form of federated content discovery.

Because the company with the de facto monopoly on web search is the same company that bought the company with the de facto monopoly on user generated video content.

Yes, but we are specifically discussing boycotting those.