The way I understand it, these are internal key pairs for the device only. You’d be publishing their pubkeys with your normal nostr keys, so you’d find the other devices pub key cause it would be an event signed by you.

Device A knows about device B as soon as A sees another public encryption key in a Nostr event signed by the same user. So A and B do *not* have a different Nostr keypair. This is about syncing encrypted data on different devices for the same user I believe.

Right, I assumed that must be it. The title of the diagram, inferring that the diagram itself deals with identity keys, is probably why several people (incl. me) were confused?