Steward

In testing

Monitoring is good at noticing that something has broken and useless at doing anything about it. Steward takes it from there: each kind of failure has a written procedure, and Steward is what carries it out. It works out what it would do before it does anything, declines to act at all unless it has been explicitly allowed to, and can be stopped dead with one switch -- because an automatic fixer that gets it wrong is worse than the fault it was sent to fix. Deployed, but not yet leaned on hard enough for us to claim it works.