Forward-looking process mining enables organizations to anticipate future process behavior and support decision-making beyond retrospective analysis by using simulation to model alternative scenarios and perform what-if analysis. Existing simulation approaches, however, are often limited to fine-grained event logs. They rely on ad hoc assumptions, and lack integration across different levels of analysis. This paper presents recent advances in the mining of simulation-based processes and places them within a unified research framework. We highlight key gaps, including the absence of systematic design principles for data-driven business process simulation, the neglect of aggregated perspectives, and the lack of interplay between coarse- and fine-grained models, and propose methods to address these issues. Our contributions include a reference meta-model for simulation design, systematic transformation of event logs into coarse-grained representations, data-driven system dynamics models, hybrid simulation approaches, and tool support. We conclude with a discussion of open challenges and future research directions, intending to establish simulation as a core pillar of forward-looking process mining.