TAE guidance emphasizes that Agile/iterative delivery drives frequent change and frequent regression risk, which often leads teams to expand automated regression suites over time. As suites grow, they can become slower, costlier to maintain, and harder to keep stable—especially if the suite is concentrated too heavily at the UI level. For this reason, TAE stresses investing in automation across multiple test levels (unit/component, API/service, and selected UI), aligning with principles behind balanced automation strategies (often illustrated by the “test pyramid”). This directly supports option A. Option B is not generally true: in Waterfall/V-model, testing activities (including automation design and implementation) are planned and may start early, but execution and refinement occur across phases aligned with integration and system readiness—not “usually only during the last phase.” Option C is too absolute: the test pyramid is a common heuristic, but TAE does not mandate it “regardless of context”; constraints like legacy systems, risk, architecture, and tooling can change the optimal distribution. Option D is incorrect because unit testing is typically a developer responsibility in both Agile and V-model contexts; testers may support, review, or contribute but do not “write automated unit tests” as a defining V-model rule. Therefore, A best matches documented lifecycle realities and maintenance concerns.