Silent Data Corruptions (SDCs) pose a significant and often hidden threat to the reliability of large-scale computing infrastructure, as they can silently compromise data integrity without immediate detection. Detecting such behaviors at hyper-scale is often challenging due to their intermittent nature and the vast diversity of hardware and workloads in a large-scale fleet. This work addresses these challenges by introducing PinDrop, a characterization methodology which leverages continuous, high-frequency testing infrastructure across millions of servers to gather information about SDCs at scale. Leveraging extensive test suites tailored to mimic real-world applications and exercise a wide range of CPU features, we provide the most comprehensive characterization of SDC failures to date, analyzing over 500 million test executions across millions of devices. Our findings reveal that 0.035% of tested machines suffer from at least one SDC failure during their lifetime. Examining years of data (rather than just a testing snapshot in time), we observe SDCs emerging long after initial deployment and persisting over time. Detailed analysis shows that an average of 0.0024% of tested machines begin failing in each quarter they are tested beyond an initial burn-in period, confirming a critical need for continuous testing. Our findings also provide further insights into detailed SDC behaviours observed, including failure breakdowns across architectures, test families, specific core IDs, and output-level behaviours.