We should probably move the restriction on cv-qualifiers of the member function vs. the object expression to [expr.ref], where the expression is actually formed. @zygoloid , any opinion here?

The restriction on cv-qualifiers of the member function vs the object expression are already enforced in two other places: in overload resolution and in [expr.call]/7 (which says "the this parameter of the function is initialized with a pointer to the object of the call, converted as if by an explicit type conversion").

I wouldn't want to move this wording to [expr.ref] unless we can remove it from [expr.call], which I don't think we can, since we do still need to define the initialization of the implicit object parameter in [expr.call].

Curiously, the pointer-to-member case currently says it has the same restrictions on cv-qualification that are described in [expr.ref], which doesn't describe any such restrictions...