Testing is complete when the required level of confidence has been achieved, not when arbitrary conditions are met. Time running out (option A) is a constraint, not a completion criterion. Finding no more faults (option C) is unrealistic and doesn't prove sufficient coverage. Option D is subjective and unmeasurable. Option B is the only criterion based on measurable confidence levels and risk assessment.