simplex are designing it wrong.
nostr messages are all signed, they only need checking, other than that, identity is ironclad. so this changes how you can do a distributed database if the events are sovereign.
the question is not about whether the events are authentic, but about whether you care to forward them on, or see them in the first place.
nostr can use an eventual consistency consensus that is not probabalistic like pBFT and Nakamoto Consensus relays can produce a consensus mostly by gossiping.
in actual fact, to some extent nostr is already running various kinds of simple, some even retarded, gossip chatter propagation patterns. these aren't a consensus but they are 50% of what you need to make one.
add the actual consensus algorithm that causes nodes to tend towards having the same view of the network and you can rely on having the same answer from a query anywhere that has a subscription to sync what you want to find.
the distinction is huge. it means with a nostr like event protocol you can do arbitrary consensus, not only a single global database view.
i will be demonstrating this in the coming months. first two use cases that it can apply to are a name registry and broadcast automation.
with a consensus running over many nostr relays we can have our own DNS registry system, even, there can be relays that specialise in this registry. when i say "broadcast automation" i mean making it so the sync workers of relays know who wants what so they just push it to them.
it can be so much more powerful and censorship resistant this way. when you use blockchain style single global databases you massively raise the baseline cost of participation. same principle in play why the nakamoto consensus stays decentralised where pBFTs rapidly aggregate into oligarchies of 3-5 individuals who are usually serial blockchain pirates.