While the accident rate of airplanes has decreased over decades, accidents still occur mainly due to unexpected failure modes. The objective of this paper is to model in the probabilistic design framework how unexpected failure modes affect reliability, and how post-design tests, called the Building-Block tests, contribute to discovering them. The probabilistic design approach can provide designers and decision-makers with stochastic insight on reliability, i.e., probability of failure. Designers model epistemic and aleatory uncertainties by assuming that they follow certain distributions. However, the current approach does not include the effect of unexpected failure modes, which is the major contributor to most accidents. In the paper, unexpected failure modes are modeled using a large error in epistemic uncertainty, and hierarchical design and test processes to detect them are proposed, as well as a simulation procedure to calculate the associated probability of failure. As an example problem, it is shown that ignoring the effect of unexpected failure modes yields an unconservative estimate of probability of failure by orders of magnitude. It is also shown that the Building-Block test under ultimate load conditions compensates significantly for such unexpected modes, and is effective for discovering modes associated with large errors.