The crux of this paper is that floating point expressions can be
evaluated in a wider format, but the return statement in a function
should still return a value of the function's return type rather than
the wider format type.

Note, this is an Annex F conformance requirement and we do not
currently claim conformance to Annex F, so technically we conform
either way.