Currently many users have trouble enabling MyISAM auto repair feauture as it introduces hard dilema: - if you will use small myisam sort buffer repair speed will be very slow and so downtime will be larger than it could be. - If you use large myisam sort buffer MySQL may crash once again because very large amount of tables may start to repair at the same time. Moreover having too large number of repairs going in parallel is not effective anyway as it will only change sequential disk access to random-like one without any improvement. As solution variable has to be added which would limit number of repairs (any repairs) going in parallel. Normally user would set this to number of CPUs as process is often CPU bound. Replication routine has to be changed to obey this variable setting.