Have you seen the relay menu on #Lume?

It has a huge list of global relays + relays used by followers.

If the current optimal number of relays to use is 2-5, I think such list make a lot of sense, but could perhaps use extra metrics to guide the user towards their usage.

Example, metrics such as latency (Again, just using Lume as an example, but that seems to be missing), whether it's paid, etc

I think another problem isn't even the centralisation at the user level. But rather, at the ISP level.

Arguably, well-operated relays that are not hosted by a cloud provider are better in every sense (assuming bandwidth needs are met by whichever ISP is used).

I think this is the next best step after self-hosting relays.

Integrating relays into clients can also solve a lot of problems (has its own challenges).