> That’s just an invalid Unicode string, then. Unicode strings are sequences of Unicode scalar values, not code points.

Because surrogates were retrofitted onto UCS-2 to make it into UTF-8, they are both code units and (reserved) code points.