We investigated the effects of interruption frequency on interrupted and uninterrupted performance in a multi-step procedural task with sequential constraints. One trial in the multi-step task consisted of five steps, which had to be learned and executed consecutively in predefined order. The task was occasionally interrupted by a letter classification task. After completion of that interruption task, participants needed to resume to the multi-step task at the correct step. Dependent variables were response times and rates of sequence errors (deviations from the prescribed order of steps), which were measured at the step after an interruption (resumption performance) and in the uninterrupted steps. Interruption frequency varied between subjects and was manipulated twofold, first, as the percentage of interrupted trials at all (25% vs. 75%) and second, by the number of interruptions per trial (i.e., once or three times) resulting in four different groups (low: 25% of trials, each interrupted once; mid: 25%, interrupted three times; mid: 75%, interrupted once; high: 75%, interrupted three times). The results revealed that participants resumed faster after interruption in the high-frequency group compared to both 25%-frequency groups, and more accurate compared to the low frequency group. At the same time, participants of the high-frequency group performed slower in uninterrupted trials compared to all other groups. These results show that high interruption frequency in multi-step tasks induces strategic shifts in participants’ control mode and encourages a more flexible processing style causing better performance during interrupted trials for the cost of worse performance in uninterrupted trials.