Why does it usually leave you worse off? Because, since it's garbage collected, the language usually doesn't see the need to also provide a good way for the programmer to manage lifetimes, because, hey, we've got garbage collection, so why would they need to? The result is that the programmer has no help at all with the other resources.
So that's where I think the line is. Do you just have memory to worry about? Or do you have other resources as well?
And, to be clear: Yes, if you only have memory to worry about, garbage collection is fabulous. A whole set of problems just... goes away. It's very freeing.
I know Haskell is trying to get Linear Types for those purposes (with a paper + working ghc branch) but even those are going to take time to get nice imo.
> I'm guessing [types are] generally more applicable than lifetimes.
I think this is an open question! A lot of people argue (you can see some of it in this thread) that Rust's rules teach them a certain programming style that carries over into other languages.
And Haskell's types aren't necessarily what you want either (I still write bash and python and go for a reason!), but I'm guessing they're more generally applicable than lifetimes. I would love to be wrong & learn more though :)