Interruptions are an inevitable, and often negative, part of everyday life that increase both errors and the time needed to complete even menial tasks. However, existing research suggests that being given time to prepare for a pending interruption—a lag time—can mitigate some of the interruption costs. To understand better why interruption lags are effective, we present a series of three experiments in which we develop and test a novel sequential decision-making paradigm, the mazing race. We find that interruption lags were only beneficial when participants had a clear strategy for how to complete the task, allowing them to avoid specific errors. In the final experiment, we attempted to use what we learned about the kinds of errors introduced by interruptions to develop a feedback-based intervention, aimed at dealing with situations in which interruption lags are not possible. We found that feedback was, only in certain situations, an effective replacement for an interruption lag. Overall, however, because the usefulness of interruption lags depend on the specific strategy a participant adopts, developing generic interventions to replace interruption lags is likely to be difficult.