It doesn't, because I think NIP-85 is badly designed.
It also will scale terribly. The idea is that, for each user, a provider publishes N events, one for each other user in the network.
If there are 1000 users of a provider, and there are 1M users in Nostr, this means publishing and keeping updated 1B events! This is insane. Most of these ranks are not needed at time X, they just need to be computed at request time.
Now, I understand your concern about Vertex being a single provider, and that's why I'm designing a protocol consisting of simple HTTP endpoints. Think blossom for WoT. I will share soon, and will make Vertex implement it later, after gathering feedback from the community. I'll try my best not to make the protocol overfit Vertex usecases and leave room for other providers and algos
