Any time you have a character with a special meaning you have to handle that character turning up in the data you're encoding. It's inevitable. No matter what obscure character you choose, you'll have to deal with it

It's a lot easier to type comma than control characters and it's a lot easier to view comma than a tab (which might look like a space).

For automated serialization, plain text formats won out, because they are easy to implement a minimal working solution (both import and export) and more importantly, almost all systems agree on what plain text is.

We don't really have Apple formatted text, that will show up as binary on windows. Especially if you are just transferring id's and numbers, those will fall within ascii and that will work even if you are expecting unicode.

The ASCII control characters do not appear well or are editable in a plain text editor.

I did always use TSV and I think the original use of CSV could have used that.

But TSV would still have many issues.

I don't understand why CSV became a thing in the 70s when S-expressions existed since at least the 50s and are better in practically every way.