Despite growing interest in multiuser virtual reality (VR) for education, evidence-based guidelines for designing effective collaborative VR learning experiences remain limited. This scoping review analyzed 23 empirical studies of collaborative learning in head-mounted display VR environments, exploring how contextual factors and technological affordances — including collaboration modality and system symmetry — shape activity design. We identified six distinct patterns of activities and analyzed the application of Computer-Supported Collaborative Learning (CSCL) scripts to support collaboration. Findings highlight predominant use of play-level (48%) and scene-level (48%) CSCL scripts, with minimal scriptlet-level implementation. Analysis of relationships between design dimensions, activity patterns, and collaboration supports reveals three fundamental design tensions: structured scaffolding versus flexible social interaction, role asymmetry versus technological symmetry, and shared physical presence versus distributed collaboration. This work contributes empirical foundations for collaborative VR learning design, while identifying gaps, design implications, and opportunities for advancing both HCI research and educational practice in immersive environments.