Under what conditions can data races occur in Python?

Data races in Python are possible when multiple threads or processes access shared mutable data without proper synchronization. Despite the presence of the GIL (Global Interpreter Lock), which restricts the concurrent execution of Python bytecode in multiple threads, data races can occur:

• When using modules that release the GIL (e.g., when working with C extensions or I/O).
• When using multiprocessing, where processes may access shared resources (e.g., shared memory, files).
• When working with asynchronous code if access to shared data is not protected.

To prevent data races, synchronization mechanisms are used: locks, semaphores, queues, and other primitives from the threading or multiprocessing modules.