Models based on Reinforcement Learning (RL) have been combined with Discrete Event Simulation (DES) to mimic the human learning process in decision-making in production scheduling (PS) problems. By conducting a survey on the current state of the art in research involving the RL-DES-PS trinomial, this scoping review seeks to summarize the findings of scientific literature, explain similarities and dissimilarities between the articles searched, and identify gaps and potential applications. The articles retrieved for the study were analyzed and a taxonomy of 12 relevant attributes was proposed. This taxonomy groups the attributes into four categories (scheduling context, RL modelling, DES attributes and DES-RL integration), and may lay a foundation for further explorations. This paper also provides insights of the existing gaps in the literature and poses questions for future research. The majority of the articles reviewed considered simplified versions of production systems, or a more limited modeling for the RL algorithm, which reinforces the need for further development and maturation in this field to address the challenges posed by production scheduling problems. Moreover, practitioners can utilize the insights from this paper by applying the proposed taxonomy as a supporting tool for developing improved DES-RL-based models, to tackle production scheduling problems.