Worked on the TencentCloud/TencentDB-Agent-Memory repository to enhance the resilience and reliability of the memory-tencentdb provider. Developed a comprehensive self-healing and recovery unit test suite that validates core recovery workflows, including watchdog behavior, lazy probe recovery, and clean shutdown procedures. Leveraged Python, pytest, and unittest.mock to simulate supervisor and gateway states, enabling thorough testing of process resurrection and deadlock recovery without requiring real infrastructure. This approach improved test coverage and reduced production risk by ensuring that critical recovery mechanisms function as intended, while also supporting faster incident diagnosis and ongoing regression prevention through robust automated testing practices.