Is this editorial or LWG?

If the Standard disagrees with the ingress paper (P0009R18) it's editorial. In this case, they agree. That generally makes it an LWG issue to determine which one is correct, unless you can present a convincing argument to the editors.

I'll point out that both layout_left::mapping::extents and the exposition-only extents::fwd-prod-of-extents are noexcept, so computing the return value extents().fwd-prod-of-extents(i) is not potentially-throwing.

I saw that many functions that accept indices in <mdspan> have the constraints of (is_convertible_v && is_nothrow_constructible_v). I wonder why not just simply use is_nothrow_convertible_v directly, which seems to be more accurate to me.

I saw that many functions that accept indices in <mdspan> have the constraints of (is_convertible_v && is_nothrow_constructible_v). I wonder why not just simply use is_nothrow_convertible_v directly, which seems to be more accurate to me.

CC @crtrott for insights.

That said, if this is not an editorial misapplication of the approved paper, I would prefer if you filed an LWG issue. If you provide rationale and a resolution, it might get approved very quickly.