While I can see a convenience argument for the somewhat contrived case of typing an entire file out by hand, entering the character once for the import does not seem like a great bar.

> it is nice to be able to easily type it.

Again, that's where an IME is helpful; on the OSes I mention, it's "␞" is:

  macOS: ⌘+^+Space, "record", <arrows to select>, Enter
  Linux: Super+e, "record", Space, Space, <arrows>, Enter

The process is highly visual on both, so you're getting feedback about whether you're about to hit the right character, or not.

(And like, if you have the file, you can always ^C ^V the character, if say you don't know how to IME, or you don't know what the name of the char is, etc.…)

Yes, it would be possible. You'd have to make sure the character didn't appear / no escaping at all was present, which the data may or may not allow.

Might as well just get a dedicated CSV→USV converter, though.

(I have a variant of this problem in JSON→YAML … usually I just shove one into the other & pray, akin to your sed solution.)

> Anything that involves something more complicated than just pushing a single button on a keyboard (this includes IMEs)

My point is that this is merely a more stringent argument; it's now "on a keyboard, and cannot involve dead keys, etc." … which now excludes classic CSV, too, which requires two keys to enter a double quote. (Shift+')

Again, it does require more keys, and it is certainly not convenient, but n keys to me is still n keys. The real question is why one isn't using a proper tool to enter the data, and is instead encoding it by hand, which, again, even for a classic CSV, is basically something I've never done. (… because why would one?)

The fact that it is a character on the keyboard is exactly the problem, too. Any character a user can easily enter will definitely end up mixed into your data somewhere.