I treat reliability as a design requirement, not as rescue work after an outage. What interests me is how a system behaves under load and partial failure, how quickly a problem becomes visible and what it costs to get back to normal - those three things, not the number of tools in place, are what decide production stability.