If one searches for a character in CJK extension C on zh.wikipedia.org, he will get all pages containing any of the characters from CJK extension C and these characters will be displayed as U+FFFD. (For instance: https://zh.wikipedia.org/w/index.php?search=𨨏&title=Special:搜索&fulltext=1 ) It seems that all these characters are wrongly treated/indexed as U+FFFD.
Note: This definitely needs fixing, but a partial workaround is to search for the 32-bit characters with quotes around them. Then the high and low surrogates are treated as a phrase. (For instance: https://zh.wikipedia.org/w/index.php?search=%22%F0%A8%A8%8F%22&title=Special:%E6%90%9C%E7%B4%A2&fulltext=1 ) This doesn't solve the problem for all use cases, but it is helpful for some.