The tartget page's URL was not being URL-escaped when passed as a parameter into the query string. Result was that any URL-encoded bytes in the link would be unescaped and interpreted as raw UTF-8 when our bookmarklet post target page reads its parameters; this would break 8-bit high characters (eg a Latin-1 %FC turning into a corrupt UTF-8 byte) or reserved chars that needed to be kept escaped (eg a %3F turning into literal ?)
Unfortunately as this was a bug in the original bookmarklet's JS code, this'll only fix it for people who grab a fresh copy of the bookmarklet.
