Does this paragraph serve a purpose? @zygoloid asked that here and the only response was that it's indeed useless. Isn't the check on requires a == 0; already rejected by [expr.const]/5.9?

This also seems like a case that P2280 ought to address, making something like this valid:

struct A {
    constexpr int size() const { return 42; }
};

template <class T>
concept C = requires (T t) {
    requires t.size() > 10; 
};

static_assert(C<A>);  // only clang rejects, explicitly citing this rule
static_assert(C<A&>); // everyone rejects, but P2280 would presumably allow 

Isn't the check on requires a == 0; already rejected by [expr.const]/5.9?

Good point. This seems to be missing a core issue. I've asked for one.

See CWG2517

Should have been superseded by #6118 (7a947d7).