Making lots of little spot improvements to more-speech. Some relays are sending events more than once, so I short-circuit those duplicates very early in the process so that they don't require processing time. That helps to minimize the backlog when those relays are high traffic.
Discussion
Bloom filter in memory to filter duplicates?
Yeah, a simple map of ids to relay urls. If a relay sends the same id, I see the url in the map and skip any further processing. If the event comes in on another relay, then I note that in the event record in the db. That way you can see all the relays that an event came in on.
relay defense shields, on