i can speculate that they open long running subscriptions to all the relays and feed them into a giant fucking database

i literally just wrote a "firehose" thing using fiatjaf's go-nostr relay pool library a couple days ago - just set the kinds in the filter and out they come

i'm working on a multi-protocol aggregator and this is how the garbage flows from Bluesky firehoses too, nostr just doesn't have a special stupid shitcoin dev name for it