UTF-8's original specification included 5-byte and 6-byte encodings to cover the complete astral plane (31-bit code points), but later specifications have marked those "invalid" today due to the current 21-bit limit of UTF-16 and to align both specifications for now rather than fix the bugs in UTF-16 (or scratch UTF-16 altogether). In theory, UTF-8 can even extend beyond 6-byte encodings (and UTF-32 into 8-byte encodings and beyond) if the next plane (63-bit code points) or the one after that ever needed to open up. (No one expects that any time soon, of course. Today Unicode is nowhere close to in danger of filling 21-bits much less 31. That would be a massive shock and the compatibility headache would be terrible with UTF-16 breaking or today's software breaking that hard codes the assumption that UTF-8 should never go past 4-byte encodings.)

If it wouldn't surprise you then I think you should recalibrate your feelings about how surprising Unicode encodings are. There aren't very many of them, they haven't changed in a very long time, and they don't deal with any of the stuff that makes Unicode very complicated (collation, combination characters, etc). They just encode 21-bit integers, albiet sometimes in a highly convoluted way for backwards-compatibility reasons (UTF-16). It's not the kind of thing that needs to be estimated, or where a layer of FUD is warranted (as it kind of is with combination characters). When talking of codepoints, it's just "up to 4 bytes", high confidence, nothing more to it.