Abstract
As Extended Reality (XR) transitions from single-user experiences to persistent, multi-user social environments, the complexity of maintaining a synchronized shared reality increases exponentially. This paper presents the first large-scale empirical study of software defects in multi-user XR systems. By analyzing over 2,600 real-world bug reports from diverse developer and user communities, we categorize the unique symptoms, root causes, and consequences of failures in these systems. Our findings highlight the "presence tax"—where technical glitches fundamentally break social identity—and identify critical vulnerabilities during session transitions and network synchronization.
Introduction
Multi-user Extended Reality (XR) encompasses Virtual, Augmented, and Mixed Reality systems where multiple participants interact within a shared digital space. Unlike traditional multiplayer gaming, multi-user XR requires high-fidelity, low-latency synchronization to maintain a "sense of presence"—the psychological feeling of being physically present with others. When synchronization fails, the consequences extend beyond mere gameplay frustration; they can induce motion sickness (cybersickness) and destroy the social fabric of the virtual environment.
Despite the growing importance of these systems, there has been a lack of systematic understanding regarding their failure modes. This study fills that gap by providing a comprehensive taxonomy of defects, offering a roadmap for developers to build more resilient shared worlds.
Methodology
The researchers employed a multi-stage empirical approach to capture a 360-degree view of system reliability:
- Data Acquisition: A total of 55,943 reports were scraped from GitHub, developer forums (Unity, Unreal), user communities (Reddit, Stack Overflow), and major app stores (Oculus, Steam).
- Filtering and Classification: Utilizing a combination of manual review and LLM-based classification (DeepSeek v3), the dataset was refined to 2,649 unique, relevant bug reports spanning 2018 to 2025.
- Qualitative Analysis: The team used iterative open coding to develop a taxonomy across three dimensions: Symptom Manifestation (visual/auditory failures), Root Cause Origin (logic/network errors), and Consequence Severity (impact on user experience).
Taxonomy of XR Defects
Symptoms and Manifestations
The study identified that connection instability and disconnections are the most prevalent symptoms (33.5%). However, XR-specific symptoms like "T-posing" avatars or "pink error models" represent a unique category of failure that directly impacts social presence. Performance-related issues, such as lag and freezes (12.3%), are particularly hazardous in XR as they are primary drivers of motion sickness.
Root Causes
Network and Synchronization Logic (33%) and Platform/SDK Integration (27.6%) emerged as the dominant root causes. Many defects arise from developers applying single-user logic to distributed environments, leading to race conditions during object ownership transfers or state updates.
Key Insights
The Fragility of Transitions
A critical finding is that 30.8% of symptoms occur during session state changes—specifically when users join or leave a room, or when the ownership of a shared object changes hands. These "transition states" are the most vulnerable points in the multi-user XR lifecycle, often leading to cascading failures where a minor desynchronization eventually triggers a system crash.
Social and Safety Implications
Beyond technical failures, the study uncovered unique risks:
- Social Identity Erasure: Avatar glitches (e.g., invisible models) strip users of their digital identity.
- Control Hijacking: A severe safety risk where one user's camera view is inadvertently controlled by another participant.
- Privacy Exposure: Behavioral data, such as precise movement patterns or avatar height, can be leaked through unoptimized network packets.
Conclusion and Future Work
Multi-user XR systems are currently plagued by "temporal fragility," where the timing of network events often breaks the shared reality. Over 34% of analyzed bugs lead to critical failures, making the application unusable.
To address these challenges, the authors advocate for the development of distributed state visualizers to help developers debug synchronization in real-time. Furthermore, "health-aware" systems are needed—applications that can detect performance drops and automatically simplify non-essential visual elements to prioritize user comfort and prevent cybersickness. As XR continues to evolve, shifting from "functional" to "resilient" shared worlds will be paramount for widespread adoption.
Summary updated