I think that nostr's most problematic technical flaw is that the "content" field in an event is a string, and not an object. Really bugs me. It's a serious limitation, and embedding stringified JSON like it's done for kind 0 events is inferior to having a native JSON object. I don't see any justification for using a simple string (although, sure, it makes things easy for simple events like short notes and reactions).
NIP-68 (picture events) really highlights this problem with these questionable "imeta" tags. Feels like the authors were like "hey where should we put this stuff ?". If "content" was an object, the content would contain a list of objects containing the url and dimension for each picture ... Is each picture's metadata stored in a imeta tag so that it can be queried with relay filters ? (like on the picture's "alt" description).
Nevermind though, i really like #nostr.