Yep.
We were recently storing tokens in a database, and I chose to use SOH for the metadata and SOX for the text.
One byte width, no collisions with printable text, and that's what they're there for.
I'd love to see a CSV replacement that used SOH ... SOX for headers, RS as "commas", and GS as "newlines". You'd be able to cleanly concatenate multiple files, since the first line is no longer special, and you'd be able to have commas, newlines, and in fact any printable text whatsoever inside the data.
And the semantics are perfectly clean. Again, that's what they're there for. Some small challenges for hand-editing that a competent text editor can easily rise to.