The wording in subclause 7.16.1.1 of both C17 and C23 has

"[if] type is not compatible with the type of the actual next argument (as promoted according to the default argument promotions) [...]".

[cstdarg.syn] only clarifies that the default argument promotions are replaced with the C++ version, seemingly leaving the C definition of "compatible" as operative for C++ in this context.

Consider, for example, the topic of CWG 3046, "Enumerations as part of the common initial sequence": Enumerations with the same underlying type need not have equivalent sets of values in C++, whereas C enumerations with the same underlying type are compatible and capable of representing the same values. Thus, enumerations should only be defined by C++ as being layout-compatible if, in addition to having the same underlying type, they also have equivalent sets of values. This difference between C++'s "layout-compatible" types and C's "compatible" types means that we should avoid using C's "compatible" in C++ contexts.

Furthermore, C23 expanded the bullets.

It now reads:

[…] except for the following cases:

• both types are pointers to qualified or unqualified versions of compatible types;

• one type is compatible with a signed integer type, the other type is compatible with the corresponding unsigned integer type, and the value is representable in both types;

• one type is pointer to qualified or unqualified void and the other is a pointer to a qualified or unqualified character type;

• or, the type of the next argument is nullptr_t and type is a pointer type that has the same representation and alignment requirements as a pointer to a character type.

Assuming that "compatible" is read as "layout-compatible", the first bullet, whose effect in C is to allow limited differences in qualification, allows further differences in the pointee types in C++ because pointers are not layout-compatible in C++ when the pointee types are layout-compatible but not the same type.

Clearer wording, likely with more restrictions than C, is needed. A question, though, is whether "compatible" should generally be replaced (for `va_arg`) with "layout-compatible" or a requirement for being the same type. In the specific case of the first bullet, I believe the latter is the right choice.