What I'm describing is a situation where users themselves have to proactively subscribe to a connection using some sort of out-of-band mechanism. For example, if a website wanted to send you emails, they could produce some sort of "connection ticket" that you can give to your email client in order to subscribe to them.
This would result in half the planet being frustrated all the time and the other half never getting their mail.
If your goal is to secretly destroy email this is the way.
Yup. I do want people to be able to contact me regarding my homepage. It's only a niche page on a niche subject, so only a handful of people has written in, but it was nice hearing from them and some of them did make quite a few valuable contributions.
Some minimal obfuscation seems to be enough to keep mail harvesters away, and beyond that those mails go through the same spam filter as all my other mail traffic. Putting up a contact form would definitively be more of a hassle than just a simple mailto:-link, and then I would additionally have to start worrying about how to keep the bots away from that contact form.
similarly, any "hello pls add me to your allow-list" emails could be made auto-disappear to the "will be deleted in 30 days" folder in ~10-15 minutes, so even if you get a 100 spam messages per day you only see the last of those, you can easily pick what you are looking for, and don't worry about the rest, they'll just disappear.
(and you still have 30 days to look for messages that might be interesting/important/etc.)
...
the real missing piece is the feedback mechanism. DMARC is meh. of course large senders have implemented FBL, but they are not available for mere mortals.
https://en.wikipedia.org/wiki/Feedback_loop_(email)