<p>Group scheduling, a classic problem in manufacturing, has been extensively investigated over the past few decades. Nonetheless, existing research rarely integrates the interactions among three key factors (setup times, deterioration effects, and learning effects). This could bring about suboptimal scheduling outcomes because these factors not only coexist but also interact dynamically in real-world production environments. In this study, a parallel machine group scheduling problem characterized by time-dependent setup times, where job processing times are simultaneously affected by both deterioration and learning effects, is explored to minimize the makespan. Specifically, the single-machine case is first analyzed regarding its structural properties. Subsequently, a polynomial-time algorithm is developed and theoretically proven to yield optimal solutions. Concerning the more complex parallel machine scenario, a hybrid metaheuristic algorithm, called CSVNS-IG, is established by integrating Contrastive Shaking-based Variable Neighborhood Search (CSVNS) with an Iterated Greedy (IG) strategy. Within this framework, the Contrastive Shaking (CS) operator and three distinct neighborhood structures (NS) form the global exploration component, enabling broad coverage of the solution space to discover diverse candidate solutions. Meanwhile, the Iterated Greedy (IG) strategy, serving as the local exploitation component, refines these candidates to enhance solution quality. Computational experiments demonstrate that the proposed CSVNS-IG algorithm significantly outperforms several existing optimization methods, including Biogeography-Based Optimization (BBO), Hybrid Water Cycle Algorithm (HWCA), and Simplified Co-evolutionary Discrete Differential Evolution Algorithm (SCDDEA), in solving this complex scheduling problem across three key metrics of optimization capability, robustness, and runtime efficiency.</p>