Run in Any Order

Bache’s rename of Beck’s Isolated, kept here because it names the observable requirement: the failure of one test can never make another test look broken. Each test builds its own fresh fixture; results don’t depend on what ran before. Seven authorities say the same thing in different words: “Independent” (Meszaros, FIRST), “hermetic” (Google).

Distinct concern: whether a test uses real collaborators or doubles. That’s the solitary/sociable axis (Jay Fields, via Fowler), not this property, and conflating the two is how “isolated” became ambiguous.

Part of Test Desiderata.

Measuring: run the suite in randomized order (pytest-randomly, JUnit 5 random ordering, --gtest_shuffle); zero order-dependent failures across repeated permutations.