aws scaled up, then assumed pending instances were healthy, then took the healthy instances offline, only leaving pending ones. wtf.