Once when it went from the old forum code to the new one when it changed to ISO-8859-1 (which is for Western text). Oddly that worked for most people for new posts. However some (apparently non-Windows &/or non-IME users) were disadvantaged.
Twice when it switched to JIS - which was a disaster. As I'd noted before hand there are quite a few single-character bytes that are used in the double-character JIS encoded text that are mangled by the forum code (& " < > that I know of). Due to the way JIS works a single error can mangle an entire thread.
Three times when it switched to the (current) x-sjis, which almost works - except for some characters (窶?, 窶?, 窶「*, and one of the two from 窶「テ冷?ケ* that I know of). Luckily the mangling only affects one character.