I don't think I am a fan of this, as most of the time a buffer is more
immediately useful.  Perhaps João's initial change would be better in
that case, for the sake of backwards compatibility?  Or does it make
sense to mention this as an incompatible lisp change?

>                                            So it sounds like code
> written for Emacs 28 should still work.  What is missing here that
> explains the breakage?

-- 
Philip Kaludercic