Note: Even if we compare them respectively we should add that a comparison of the template parameter lists should also be required.

One more point is that there appears to be implementation divergence due to the existing wording:

template <typename>
struct Result { };

struct Base {
  template <typename T>
  Result<T> f(int) const;
};

struct Derived : Base {
  template <typename T, typename>
  Result<T> f(int) const;

  using Base::f;
};

struct S { };

void f() {
    Derived d;
    d.f<S>(0);
}

Clang will reject this code while gcc accepts: https://godbolt.org/z/Bu7h9z

The purpose of this rule is to avoid "using Base::f" to produce a conflicting declaration, and we poorly mimic the rules from [over.load] for that. For example, if the function from the base class has no ref-qualifier, but the (otherwise identical) function in the derived class has a ref-qualifier, is that ok? After all, [over.load] says such two declarations cannot appear in the a single class definition.

I'm not sure there is an editorial answer to that question.

Editorial meeting: Not editorial, send to CWG.