This paper revisits the well-known opacity caused by the interaction of post-obstruent tensification and coda cluster simplification in Korean and suggests a new class of data that threatens the validity of previous approaches. The new data shows that the opacity occurs only if the input belongs to a certain morphological category such as verb. Therefore, it calls for a theory in which morphology and phonology are systematically interleaved, such as Stratal OT (Kiparsky, 2000). I show that the Stratal OT approach provides a solution to the problem since it adds derivational effects as well as morphological insights to an OT grammar.