I agree with the above points that this debate would have to be on a case by case basis. If there is a failure in V&V testing, it is primarily important to make sure that your testing method itself is valid for testing this case. If the test is invalid, a failure or a pass would mean nothing to the protocol. Once your test method is validated, the failure must be evaluated if it is an isolated event or if it is indirectly affected by another factor of the device. If the change can be tested in isolation, then a regression test may be sufficient to save on time and cost. However, there are instances when the whole protocol must be carried out once again, while it may be time consuming, may be the only way to produce valid results, such as having many integrated parts that do not work in isolation.