Can clock faults be detected through functional test?

We analyze the probability to detect clock faults indirectly through conventional functional testing by considering realistic datapaths derived from ISCAS'85 benchmarks. We show that, even optimistically assuming that we are able to test all short and long paths for min and max delay violations, the detection of clock faults can not be guaranteed, thus mandating new, specific testing approaches for clock faults, otherwise possibly compromising the system correct operation in the field, with dramatic consequences on product quality and defect level