Automatic algorithm configuration is concerned with finding the best hyper-parameter values for some specific algorithm. These values are then fixed throughout the execution of the algorithm. Another approach is parameter control where the values adaptively change during execution. In this work, we explore the hybrid concept of multi-configurations where values are still optimized before-hand, but as different sets of configurations that are then used one at a time during execution. In particular we explore a number of strategies based on fixed sequences of configurations and roulette wheel selection, and compare them to some baselines. We evaluate the strategies in the context of Iterated Local Search on the Permutation Flowshop Problem. Results show that both fixed and roulette strategies are better than the baselines, but also that roulette outperforms the fixed approach when hyper-parameters are optimized on more diverse sets of instances. We observe that the chosen values are not necessarily ones that would be considered the best in the literature because they are used as part of the multi-configurations.