I haven’t used Excel in a long time, so hadn’t checked. It seems that this is a long-standing issue with Excel, that it assumes MacRoman as the character encoding.
Checking online, people suggested that the Import Wizard would work. If you do that (create an empty spreadsheet, then Import from the File menu), you get the option to say it’s UTF-8 and comma-separated, but then it gets it wrong with the rows that contain a carriage return (U+000D) or line feed (U+000A), despite them being quoted correctly.
There’s no good solution to this problem, as far as I know. One would be to change the actual output to some other representation, such as XML or code point format, though that sort of breaks the model. If anyone has a better idea, let me know!
It does also work correctly in LibreOffice, which shows the options when you open the file. Numbers is still the easiest way to use it.
John