How do you evaluate the risk of a new, unproven software feature being introduced into the flight stack?
