Yeah, nostr:npub1useke4f9maul5nf67dj0m9sq6jcsmnjzzk4ycvldwl4qss35fvgqjdk5ks was talking about that. It'd be a simple script, really, so long as the document had only 2 levels: header and content-articles (which most would). 

The 30040 contains the ePub metadata and the list of source files for the document, then you just have to find all the json files on the relays, extract them all into one markdown file, and convert them with pandoc.

And the breakdown would just go the other way around. We're actually usually starting with one markdown file, so that's actually how we're doing it, now, using a script to split it up into 30040 header and 30041 sub-events.