To reach Europe’s net-zero targets by 2050, the transportation sector must become more energy efficient. Lightweighting is key, and polymer composites are appealing because they enable substantial weight reductions without compromising performance. Manufacturing methods such as overmolding enable integrated, highly optimized parts with superior mechanical properties. Yet industrial uptake is held back because their mechanical behavior is difficult to predict. A major challenge is predicting the long-term response of polymer composite structures. Methods that perform well for standard fracture specimens do not straightforwardly transfer to components with more complex geometries. For simple geometries with a prescribed initial delamination front, Paris-type cohesive zone models can describe delamination growth under high-cycle loading. However, in more complex configurations, such as quasi-isotropic laminates, multiple mechanisms—including intra- and interlaminar cracking—develop simultaneously. Under these conditions, Paris-type formulations typically cannot capture both fatigue initiation and subsequent growth. The challenge becomes even more demanding for thermoplastic composites, where viscoplastic deformations accumulate during cycling, altering local stress ratios and fatigue damage rates. In this contribution, we present a modeling framework that addresses these limitations. The framework combines a cycle-dependent cohesive zone model covering initiation and propagation with a time-homogenized viscoplasticity model for continuous fiber-reinforced thermoplastic composites. The approach is demonstrated on two cases: multidirectional laminates and overmolded thermoplastic composites. For multidirectional laminates, transverse cracking and its interaction with delamination are captured using a cohesive XFEM formulation. For overmolded composites, the framework is applied to a rib pull-off test on a T-joint extracted from an overmolded thermoplastic panel. The simulations provide insights into the effects of viscoplastic deformation, processing-induced mesoscopic geometries, and boundary conditions. Finally, the sensitivity of key modeling assumptions and processing parameters is assessed, supporting both process optimization and more reliable structural design of polymer composite components.