The developers behind Scuttlebutt [0] should apply for this! It pretty much perfectly meets Mozilla's criteria.
I think that the real solution requires caching, lookups, and multiple types of access through fingerprinted (sorry, can't say "secure") forwarding and proxying. It's a big task, but I do think that a small example could be built within the next year.
Heck, even LoRa, with the right software behind it could be a contender, if married to the correct technology.
Hm. That gets me thinking.