Follow up on comments in ImplicitConversions_07 unit-test

[AlekseyTs]

        [Fact]
        public void ImplicitConversions_07()
        {
            var source =
@"class A<T>
{
}
class B<T>
{
    public static implicit operator A<T>(B<T> b) => throw null;
}
class C
{
    static B<T> F<T>(T t) => throw null;
    static void G(A<object?> a) => throw null;
    static void Main(object? x)
    {
        var y = F(x);
        G(y); // warning
        if (x == null) return;
        var z =  F(x);
        G(z); // ok
    }
}";
            var comp = CreateCompilation(new[] { source, NonNullTypesTrue, NonNullTypesAttributesDefinition });
            // PROTOTYPE(NullableReferenceTypes): Several issues with implicit user-defined conversions and
            // nested nullability: should report `'A<object?>' doesn't match ... 'A<object>'` rather than
            // `'A<object>' doesn't match ... 'A<object?>'`; should report warning for `G(y)` only, not `G(z)`
            // (see NullabilityWalker.ApplyConversion).
            comp.VerifyDiagnostics(
                // (15,11): warning CS8620: Nullability of reference types in argument of type 'A<object>' doesn't match target type 'A<object?>' for parameter 'a' in 'void C.G(A<object?> a)'.
                //         G(y); // warning
                Diagnostic(ErrorCode.WRN_NullabilityMismatchInArgument, "y").WithArguments("A<object>", "A<object?>", "a", "void C.G(A<object?> a)").WithLocation(15, 11),
                // (18,11): warning CS8620: Nullability of reference types in argument of type 'A<object>' doesn't match target type 'A<object?>' for parameter 'a' in 'void C.G(A<object?> a)'.
                //         G(z); // ok
                Diagnostic(ErrorCode.WRN_NullabilityMismatchInArgument, "z").WithArguments("A<object>", "A<object?>", "a", "void C.G(A<object?> a)").WithLocation(18, 11));
        }

[AlekseyTs]

I am not sure if the expectations stated in the comment are correct. It feels like the warning reported for G(z) is correct and there should be no warning for G(y).

[jaredpar]

Looks like this is being tracked by #29605. The code in NullableWalker uses that issue when referring to this test case.