I see this as one of the many manifestations of possibly the biggest shortcoming in the original design of the Internet's architecture, i.e. not having an "identity layer" taking care of user authentication and information ...
You could still accept email from strangers, if you wanted, and it would be no worse than today. What a standard identification system would do is that it would allow you to attribute and share reputation correctly, so it would be much easier to avoid false positives (because you can be sure that email that claims to come from your friend actually comes from your friend) and it would be easier to prevent your spam filters from attributing bad reputation to the wrong entities, making them more reliable.
We have several decades of S/MIME and PGP failing because nobody knows how to do key distribution at scale.
DKIM provides a reliable identifier per domain. In practice that's pretty close to identifying mail addresses since it is quite rare to have one party impersonate another within the same domain. DKIM Is indeed useful as an input to spam filtering. How different would a per-user identifier be?
If you mean an identifier that follows you when change e-mail addresses, we have PGP, and we know how well that works, particularly when you goof and lose your private key.
R's, John
