Which statement best summarises how transformed code should be validated?

Generate and run tests that compare transformed behaviour against expected legacy behaviour across representative and edge-case inputs, using review as a complement rather than a substitute. Only executed comparison can demonstrate behavioural equivalence, and edge coverage makes it meaningful. Thoroughly reviewing generated code and confirming the service starts cleanly is insufficient because reading and starting do not exercise business logic. Comparing application screens before and after transformation is insufficient because it misses data effects that matter. Deferring testing until the transformed service has run in production for a settling period moves defect discovery into production, where defects cost most. For the exam, state the method and the coverage together.