I’ve written thousands of lines of copyleft code. I don’t see a practical benefit if you don’t want to obligate users of that code.
As long as you consider developers of derived software to be your users — permissive makes most sense.
But if you consider end-users of software it’s definitely copyleft.
Some would argue that if you want that, then you need to license it in a way that protects anyone's freedoms and disallows other ones from stepping on them.
And whether it works or not is mostly untested in the legal sense but the idea with copyleft licensing is that big money shouldn't be able to ratfuck a free software embrace/extinguish-style, or some other way. I.e. it is supposed to protect the rights of the users to enjoy free software.
In an argument about how much someone is able to use something, I think you could argue that a copyleft license turns more people away than it would force to participate, so the overall amount of work that is freely available is less than it would have been with a permissive license. (That is, if 10% using a permissive license don't contribute back and 90% do, you end up with more contributions than if 100% of users contribute back to copyleft but you only have 50% as many users.)
I do think it's very situational though - clearly Linux being GPL has been a huge boon in forcing contributions, and I think the killer bit here is that forcing those contributions means that people are able to use their hardware that might otherwise only run proprietary operating systems. That is, it's the forcing open of something that hardware companies would otherwise keep closed.
On the other hand, I don't see any huge reason why a good regex library, for instance, shouldn't be permissive - making a regex library copyleft isn't going to force open some proprietary software, they'll just use some other (maybe worse, maybe equivalent) regex library, and you might get fewer overall contributions and end up with a worse overall available library.