Please see bug 4635984. This is the second part of the optimization where we create a desCipher once per context and use it thereafter for every message token.