• ILP model with block consistency, idle time, setup dependant times & parallelism. • Proven NP-hardness justifies metaheuristic despite restrictive constraints. • Divide-and-Conquer ILS outperforms global approaches even with stricter constraints. • Comprehensive benchmark and sensitivity analysis establish robustness and generality. • Real-data validation highlights potential to improve current practice.
Chouraqui et al. (2026) studied this question.