|Don't ask to ask, just ask|
there are several very good reasons not to put "use" into a subroutine (or other block)
I was pretty convinced already, but your list settles the matter for me.
That said, for those not convinced that every abstraction is best served by being OOified, the notion of block-scoped imports is tantalising. Makes me wonder if a new pragma--say: uselocal Some::Module qw[ :stuff ]; wouldn't be possible.
At runtime, the import list would be processed per use, but the callers package would be localised (and self initialised) first so that imports would disappear at close of block.
The problem is how to localise a hash entry within the scope of the caller?
BTW: I love your tagline. It's duality reminds of the old standby "this statement is false". Comtemplating it brings to mind the nightmares I had as a kid when I realised the implications of the infinity of space.
Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
"Science is about questioning the status quo. Questioning authority".
In the absence of evidence, opinion is indistinguishable from prejudice.