How do we achieve that? Releasing often and in smaller increments. A release causing an outage should be easier to test than the subtle bugs that come creeping

How do we achieve that? Releasing often and in smaller increments. A release causing an outage should be easier to test than the subtle bugs that come creeping one, two, three days later. Either way, if there is a catastrophic failure, its better to have everyone readily available.