Quality assurance for shipped games is closely linked to user perception and commercial performance. Even games that are positively received from a design or artistic perspective may suffer from reduced adoption if technical issues negatively affect the player experience. In this paper, we study a player-driven QA system deployed in a commercial game played by over 132,000 unique users and analyse 8,002 production-relevant feedback reports collected over 3.5 years. Our approach transforms multilingual user feedback into a unified representation and defines operational metrics for environment heterogeneity, report richness, reproducibility cues, and actionability. The analysed reports span 91 countries, 277 GPU models, 598 CPU models, and 47 game versions, demonstrating broad runtime coverage across heterogeneous environments. However, only 54.7% of reports contain explicit reproduction cues, indicating a substantial gap between user-reported issues and reproducible technical problems. Report richness is strongly associated with reproduction-cue rate, which increases from 18.3% in the lowest richness bin to 75.0% in the highest. These findings indicate that player-driven QA is an effective mechanism for post-launch product improvement, while also highlighting the need for structured feedback collection and hybrid telemetry-assisted approaches to improve reproducibility.