1 in 20 deploys has a false negative (IE, something went out without working). Continous Integration is mostly about regression, so the new tests have less val

1 in 20 deploys has a false negative (IE, something went out without working). Continous Integration is mostly about regression, so the new tests have less value than running old tests. We tend to fix issues in production (bugs) by mandating a unit test with the failing issue to be written as part of the fix.