The month highlighted in light gray:

Probably not.

Sorry, this was ambiguous: I was specifically referring to the "grey month" version. We should form a wider consensus whether we want any color-based highlighting here at all.

I don't think the colors I chose do the proposal any justice. They look like hyperlinks, one unused and one used. Using bold didn't help; the code hardly looked any different.

I find this looks much nicer:

My suggestions are a bit hideous. It'd be better to base the issue's resolution by referencing existing documents. This is probably clash with general code highlighting anyways.

I have previously prepared a PR that I did not push, where I used digit separators for all the long literals in the standard. In that case, these macros were rendered as:
#define __cpp_lib_addressof_constexpr. 2016'03L // also in <memory>

Is that something I should consider submitting instead?

The main intent in that PR was making the large numbers specified in the random numbers clause more readable.

Is that something I should consider submitting instead?

No, see:

This is a follow up for #5109.

Editorial meeting decision: we prefer if we added digit separators, and that is an observable change that requires a (small) paper (with an Annex C change). Such a paper is warmly invited.

If ' were used to split dates, this idiom would stop working.

Why? The preprocessor understands digit separators and they don't affect the value.

You're right: https://compiler-explorer.com/z/ozK9xE5qs!
I didn't think that through enough.

The breaking change is that stringification observes the change.