A race condition is a concurrency defect in which a result depends on the timing or interleaving of competing operations. It commonly appears when multithreaded code accesses shared state and fails only under a particular workload or message ordering. section 4.2.1
A failure that passes ordinary tests
An application uses the .NET Task Parallel Library, Parallel.ForEach(), and locks around shared state. Local testing and unit tests pass, but production later exposes a bug caused by a race condition triggered by a specific load and ordering of messages. The eventual response is to return to a single-threaded model. section 4.2.1
Race conditions are especially difficult because:
- They may remain invisible during local testing and appear only under particular timing conditions. section 4.2.1
- They arise alongside shared mutable state, making concurrency design and state ownership central concerns. section 4.2.1
