I take "every character" to mean "anything that was represented in a reasonably common pre-unicode code page or character encoding, as well as anything that might come up in OCR output of text documents".

Emojis obviously got in from Japanese character encodings, and imho the world is off better for that. Though many of the extensions of the emoji set really don't seem to get what emojis are used for. Similarly, chess and shogi pieces as well as symbols from Western playing cards go in through previous encodings, and domino tiles got accepted based on being conceptually similar. A bit questionable imho.

On the other hand the Azimuth sign seems to satifsy the "would appear in OCR scans", based on being published in font catalogues. Even if nobody has come forward with a book it appears in, I don't think they made and advertised lead type characters for fun. It has to have had some use in printed publications of some type (probably scientific, from the surrounding context)