Why is test isolation important and how do you achieve it?

An isolated test has no dependency on what ran before it. That property lets Jest reorder or parallelise files, lets you run a single test while debugging, and makes failures reproducible. Shared mutable state breaks all three.

Common leaks

• A spy installed in one test still active in the next
• A module-level cache populated by an earlier case
• Fake timers left enabled after a test
• Rows written to a shared database without cleanup
• A global object such as window.fetch replaced and never restored

afterEach(() => {
  jest.restoreAllMocks();
  jest.useRealTimers();
  delete (globalThis as { fetch?: unknown }).fetch;
});

test("fetches the profile", async () => {
  globalThis.fetch = jest.fn().mockResolvedValue({
    ok: true,
    json: async () => ({ id: "u1" }),
  });

  await expect(loadProfile("u1")).resolves.toEqual({ id: "u1" });
});

Prefer creating state inside the test or in a beforeEach scoped to the block that needs it. Use beforeAll only for immutable or truly expensive resources. If two tests must share data, that is often a sign they should be one test or that the fixture needs a factory.