I gather from the discussion of P2025 that
in order for it to advance further in EWG, a complete implementation is
likely required in at least one compiler. There has not been a revision
of the paper for quite some time; although I haven't spoken to Anton
about it, I would guess that he simply felt that this task was too
complex and gave up, which is what I would have done if I were in his
situation. Implementing a core language proposal in a compiler is quite a
challenge.
And actually, I think that it simply isn't worth it. What I really want to be able to do is get explicit access to the function's return slot, but also have some safeguards that make it hard for me to accidentally access it before I've constructed it. This will require some kind of syntactic extension to the language.
P2025 seemed like an appealing idea when it was a simple tweak with no major implementation concerns. Now that it has serious implementation concerns, I see no reason not to dream of a more general solution.
As far as I'm concerned, P2025 is dead unless I see evidence to the contrary. Future proposals should simply discuss it and then move on.