gah. f was down for 5 hours after a routine database modification revealed massive innodb corruption which had somehow replicated to the slave. after splitting the cluster, our last backup was nine days old and a few million records short. eek. so a painstaking search and destroy for bad records was conducted. stupid mysql.