From Jim Thomas:

This particular issue was discussed at the 2021-12-22 CFP meeting. The following is extracted from the meeting minutes (N2910):

7. [2302] C++ liaison issue from David Olsen
   Relates to arith conversions occurring between long double and _Float64.
   David Olsen says an expression converting to long double is more consistent.
   But, the current draft says that such an expression converts to _Float64.
   It was noted that a standard type may have lesser quality arithmetic than
   an interchange type which is at odds with the fact that we really want to
   lean towards better quality arithmetic
   Jim proposed some text if we decide to change.
   It was also noted that such a change also

• affects tgmath rules for which examples need to be checked, and
• adds another special case to an already complicated set of rules.
  Motivated by Microsoft Visual C++ where long double is the same as double.

Conclusion: Recommend no change - change does not justify added complexity.

WG14 already made a decision, no need for SG22 to see this issue.
From the email thread : "Given the rationale linked above, the previous SG22 discussion and related polls recorded in WG14 N2835, and the subsequent reconfirmation by the WG14 CFP SG recorded in WG14 N2910, I no longer see a reason for SG22 to discuss this CWG issue. I agree with the direction currently indicated in the CWG issue to add an annex C entry."

EWG discussed this during the Monday AM session in Tokyo. No poll was taken, but we expect to see this again later in the week after offline discussions.

Close CWG2836 as not a defect.

Consensus