What is monkey patching, and what risks does it introduce in production codebases?

Monkey patching is the practice of modifying a module, class, or function at runtime, usually to change behavior without editing the original source. In Python, you can assign to module attributes, class attributes, or instance attributes to replace methods or add new ones. It is sometimes used in testing to mock dependencies, in frameworks like gevent to patch blocking IO, or as a quick hotfix. The risks are significant: it makes code unpredictable because behavior depends on import order and runtime state; it can break other libraries that rely on the original implementation; it is hard to debug because the source code does not reflect the actual behavior; and it can break across versions when the patched internals change. In production, monkey patching should be a last resort. Prefer subclassing, dependency injection, decorators, or configuration. If you must patch, isolate it, document it, and test it thoroughly.