The article details a JavaScript memory leak issue that affected a cloud function in a synthetic monitoring product. The problem caused out-of-memory crashes several times a day, resulting in failures during data processing. The author realized that the lodash `memoize` function was inadvertently retaining references to all previous data unless the cache was explicitly cleared. By implementing a cache-clearing mechanism every other time a test result came in, the memory leak was effectively addressed. Prior attempts to solve the problem were unsuccessful, largely because the memory leak only occurred when processing unique URLs. The author had previously increased the cloud function's memory as a stopgap measure, but without access to a heap snapshot in the production environment, identifying the root cause was challenging. Despite the fix, the business impact was minimal; the savings in CPU time were small, and the function's resilience meant that initial failures didn't significantly disrupt operations. While the fix improves memory usage in the backend, it won't dramatically change front-end performance. Users typically interact with fewer test results at a time, contrasting with the backend's need to process numerous unique URLs. The overall infrastructure is designed to handle occasional failures without major consequences.