In DART's case, the MIB determined that the first cause for its premature retirement occurred when the estimated and measured positions differed to such a degree that the software executed a computational "reset." By design, this reset caused DART to discard its estimated position and speed and restart those estimates using measurements from the primary GPS receiver.

Careful examination of the software code revealed that upon reset, the velocity measurement from the primary GPS receiver was introduced back into the software's calculations of the spacecraft's estimated position and speed. If the measured velocity had been sufficiently accurate, the calculations would have converged and resulted in correct navigational solutions. However, DART's primary GPS receiver consistently produced a measured velocity that was offset or "biased" about 0.6 meters per second from what it should have been. This had the unfortunate effect of causing the calculations, which were being performed autonomously, to once again diverge until the difference became unacceptable to the pre-programmed computer logic. Once the limit as to how much the calculations could differ was reached, the software executed another reset. As a result, this cycle of diverging calculations followed by a software reset occurred about once every three minutes throughout the mission. These continual resets caused the incorrect navigational data that prompted excessive thruster firings and the higher than expected fuel usage.

Too many pre-programmed resets to the wrong data. Oops. I guess one small thing can cause a whole mission to fail! If you read the report though the rest of the mission was pretty succesful. You would have though they would build in some sort of fail safe ground control method though when there's millions of dollars worth of equipment at stake?