Although tests such as these cannot guarantee long-term operating efficiency, they establish something of distinct value. They prove that the machine overrides very well the many disturbances that are always present, crosstalk and ambient electrical noise, line voltage fluctuations, dust on the magnetic tapes, and the like. To achieve this in so complex a device requires a lot of painstaking engineering debugging to locate and correct marginal tolerances in the system. On the first machines of a new design this debugging may require many months. Until the marginal tolerances are fairly well eliminated, long error-free runs will hardly occur. The requirement of substantial error-free running periods in an acceptance test is, therefore, a powerful incentive to thorough debugging. In fact, the real purpose of these tests is to insure that, if any intermittent or sporadic errors remain in the system, the intervals between them are at least long enough to allow useful work to be done.