It looks like we've combined the header and the class synopsis here. Given how small the header is, what's the value in splitting the two? Is the status quo unclear or confusing?

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1208r6.pdf used to look pretty much like [support.initlist] looks now, but that changed during #3037.

Yes, I see. We had to fix the hanging paragraphs, but arguably source_location does that in a sligthly more principled way.

On the flip side, source_location requires excessive nesting for such a short subclause, so I'm not sure which way is better here.