When later stages of the deployment pipeline fail, such as acceptance tests or performance tests, instead of stopping all new work, we will have developers and testers on-call who are responsible for fixing these problems immediately. They should also create new tests that run at an earlier stage in the deployment pipeline to catch any future regressions. For example, if we discover a defect in our acceptance tests, we should write a unit test to catch the problem. Similarly, if we discover a defect in exploratory testing, we should write a unit or acceptance test.
