This phrasing has another issue, too: It says "if and only if", so if t() is not an equality-preserving expression, C(t()) must be different from C(t()). That seems impossibly to guarantee.
@CaseyCarter, you wanted to open LWG issues for the remaining open concepts items. Do you have one for this already?

This wording specifically (and the similar wording in [concept.commonreference]) came up during the final editorial review, and I'll be filing an LWG issue for it. I've been waiting for the updated WD to file my list of issues, since I suspect the LWG issue management utilities will fail horribly with issues that refer to stable names that don't exist in the working draft.

EDIT: I suppose we should leave this issue open for now - to be sure it isn't dropped on the floor - and I'll come back after the corresponding LWG issue has been filed and post a reference thereto?

@CaseyCarter

since I suspect the LWG issue management utilities will fail horribly with issues that refer to stable names that don't exist in the working draft.

They'll be fine (it will just be "Section 99"). We have plenty of issues targeting stable names that no longer exist.

This is now LWG3154 "Common and CommonReference have a common defect" which LWG tentatively approved in Batavia. I believe the P/R is worded in a less confusingful manner.

LWG 3154 was adopted in San Diego.