That seems bad. There's plenty of evidence that if you ask the user "hey, do you want to continue but be insecure?" they'll blindly click "yes".
While non-browser focused protocols may still rely on clients not forbidding insecure protocols, we should absolutely be trying to move away from blindly allowing insecure protocols (or effectively allowing them, given users typically click "yes").
I'd strongly suggest failing hard, and allowing per (host, port) exceptions in the config file, rather than asking every time.
I've been a heavy Emacs user since 1997, using it as my primary editor on all platforms. (In the last year or so, I've been mixing Emacs and Visual Studio Code, because of the latter's solid support for particular language servers.)
I am completely horrified at the suggestion that it might be OK for Emacs to have have insecure TLS defaults. This would be an absolute deal-breaker for me (and would almost certainly result in Emacs being banned at work, with no objection from me).
I do not have the time to fix every Emacs install on every server to be secure with the latest TLS standards. I need to be able to trust the developers of the software I use to get this right. I do not want to worry about whether `package-install` has been affected by a MITM attack.
We all agree that Emacs should provide sensible defaults, the disagreement is what that sensible defaults consist of.