One of the irrevocable paradigms in (software) testing is: test early
There are many detail explanations about this, but nothing is as vivid as the Chicken Test Story:
When Rolls Royce was developing the RB211 Engine they came insolvent in 1971. And the simple reason for this was, that they had to spend too much money in development because they tested too late. The test which finally broke the engine was a rather simple test: The Chicken Test; A so called Birdstrike Simulator throws a bird, in this case a chicken, at high speed into the aircraft engine. This is a very common real life scenario for a plane especially at take off. I think, this test is still a requirement for certification by FAA. But back to the RB211: For this engine this test was a desaster. The titanium blades were shattered into pieces by the flying chicken.
So if the requirement of robustness of the blades would have been defined and tested early in development, a redesign could have been done much earlier and the cost of this redesign would have been much lower.
So if you think early about what could happen in late tests and how you could test this also earlier, you would probably find the need for additional requirements or refinement of existing requirement. And as you have better requirement then your development will get automagically better.
Even if you won’t see any test failing during test execution, the impact of testing on requirements and development and therefor on the end product and your budget can be huge.