Effective communication is essential in high-stress environments but stress often disrupts the flow of information and leads to miscommunication. While scenario-based training exercises are widely used, post-hoc reflection and analysis of verbal interactions remain challenging due to overlapping speech, limited analysis time, and the dynamic nature of these situations. This paper introduces ROGER, a novel visual analytics interface designed to support afteraction reviews of communication during high-stress training scenarios. Developed in collaboration with police trainers through an iterative design study, ROGER integrates emotional voice metrics, heart rate variability, and spoken language content to provide a comprehensive analysis of team communication. The system enables a flexible in-depth exploration of communication patterns through motifs—repeated sequences or content elements—including those generated by a large language model (LLM) as well as predefined ones. Our approach addresses the limitations of existing tools, which focus primarily on content summarization or voice replays without incorporating emotional and stress-related voice data. We validated the utility through interviews with police trainers and conducted a workshop with medical first responders to investigate the potential for cross-domain applicability. Our findings provide preliminary evidence that ROGER supports effective team performance analysis in diverse high-stress environments.