>Won't it all be in one long line?

Depends on whether editors start a newline for \u001E. Notepad++ doesn't. Than is definately going affect readability. Hmmm.

>Doesn't seem negligible when the editability issues with CSV were also minor.

The issue with CSV, as far as I am concerned, is the escaping. And that isn't minor from either a readability or parsing point of view.

>Wanting those characters specifically is rare, but wanting to safely store an arbitrary unicode string - maybe a filename - is very common. Parsers might invent their own escaping to handle this.

Are \u001F or \u001E even legal filename characters on any OS? These codes could turn up by chance in a blob of binary data. But it isn't intended for binary data.