Does lexing happen per code point or per grapheme?

Looking at the "Whitespace and Comments" section of the Swift Lexical Structure:

line-break → U+000A
line-break → U+000D
line-break → U+000D followed by U+000A

If a line-breaking code point is followed by one for a combining character, i.e. would be considered a single grapheme, would that be considered a line break followed by an orphaned combining character, a non-line-breaking grapheme, or a lexing error?

(Unless the Unicode standard makes it illegal for a combining character to be paired with a line-level, or higher, separator.)