Nostr isnt designed to scale at all.  Connection oriented networks dont scale very well.  But they can do a decent enough job.

Sufficiently decentralized is orthogonal to scaling and imho a red herring.  We know what scales through observation and we know what doesnt scale through observation.

Nostr was under 100 users 2 years ago.  It's done very well to get where it has, but it wont overtake the internet, and isnt designed to.  It's a very useful tool, and unlocks a lot of developer innovation and creativity.  Working as expected. 

 We should also expect nostr to "fail to scale" (referencing the Andreas Antonopolous talk) as it experiences growing pains. There are surely multiple paths forward for scalability. 

 There are indeed multiple paths forward to scalability.  But nostr hasnt really explored this yet.  There's a lot of hope that the gossip model will solve things, but no evidence that it will.