Yeah we really just need to query a relay for counts. And relays should generally implement caching, which is used to very good effect and lowers bandwidth when using an aggregator relay.

Yes the count query seems like the obvious fix here. I just haven’t seen it implemented in any client yet, no idea if any Relay even support it yet. A count query has its limitation too, you have to blindly trust the relay, and have to choose only one because you can’t deduplicate 2 different count response. Cannot verify event signature too obviously. But I agree the count query will become mandatory at some point. It just seems like this kind of query is not a relay job but more a @Vertex one, just like follow count.