How we built an async-aware Python profiler

Datadog has developed an async-aware Python profiler to address a key limitation in traditional flame graphs, which often lose the logical relationships between `asyncio` tasks by reporting them as independent stacks. By reconstructing parent-child task dependencies, the new model preserves the application's execution path, allowing developers to see how different tasks are connected. This approach also provides higher-fidelity profiles while reducing profiling overhead by more than 60%.