It's complex but not that much... It just takes a few relays as inputs, downloads all events from them, collects all references to other relays and keeps growing the list of relays to check. Then it just dedupes by ID. Takes about 1hr to generate the report with about 2.5M events.

Hmm probably is O(n) then. Guess no need to do any elaborate cross indexing. Of course, with all the zaps on comments on likes on zaps on commments on likes...it could still get silly.