> What's a "corresponding pack-index-specifier"?

I believe that by reading referred [dcl.type.pack.index] such 'corresponding' would be clear.

Other cases of using 'corresponding' in naive way exist, for example in [temp.type] p(1.x).
However, never mind about it, let me focus on your proposed text for sure very interesting.

> We could say "For a type template parameter pack T, the pack-index-specifier T...[constant-expression] denotes a unique dependent type."
> but that feels like a minor improvement only.

Yes, sure, it is a minor. Indeed my title was about uniformity, i.e., no lack of consistency.

Maybe, the follows would also work;

"For a type template parameter pack T, a pack-index-specifier denotes a unique dependent type."

I believe such a kind of wording in p5 adds quailty at zero cost, due to the uniformity with p6,

similary to how p4 discusses both topics in a single point for decltype(e).

Thanks,