I understand why this would affect collation and indices, but why can't I even store the four byte characters in an utf8mb3 table? The SQL parser and printer se

I understand why this would affect collation and indices, but why can't I even store the four byte characters in an utf8mb3 table? The SQL parser and printer seems to deal with the characters fine, because it is independent from the table. I would assume it would just stuff the UTF-8 bytes in there as they are. Is it trying to validate them? Or is it using UTF-32 (or a truncated version) internally and saving everything as 3-4 bytes?