How adding 'async def' in FastAPI can silently freeze your entire service

A common FastAPI refactor — converting plain def endpoints to async def for cleaner code — can trigger a severe performance incident where every endpoint, including lightweight health checks, slows simultaneously. The root cause lies in how FastAPI handles the two endpoint types: plain def functions are offloaded to a 40-thread worker pool, while async def functions run directly on the single-threaded event loop. When a developer adds async def without replacing blocking calls inside — such as synchronous HTTP requests or blocking database queries — those calls monopolize the event loop, stalling all other requests for the duration. Unlike Node.js, where the event loop is the danger zone, Python's FastAPI uses the threadpool as a safety net, meaning removing a function from that pool by marking it async is a downgrade, not an upgrade. The correct approach is to either keep blocking code in plain def endpoints or fully replace blocking libraries with async-compatible alternatives like httpx before switching to async def.