Test isolation guarantees that a test depends neither on execution order nor on state left behind by another: no shared mutable static field, data recreated for each test, resources cleaned up. An isolated test is reproducible and can fail on its own for a clear reason.