Fail-slow disks pose a distinct challenge due to their subtle yet insidious nature. They exhibit performance degradation that may not be immediately visible but can lead to significant slowdowns and reliability issues within large-scale storage systems. Traditional redundancy and fail-over mechanisms are designed to address outright disk failures but are less effective at detecting and mitigating the gradual performance decline associated with fail-slow disks. The two primary symptoms of fail-slow disks--consistently higher latency compared to peer disks and recurrent abnormal spikes--make it difficult to establish fixed thresholds for alerts or accurately track performance trends. In light of these challenges, there is an urgent need for advanced detection mechanisms that can proactively identify and address fail-slow conditions.