every generated test mocks the thing it is supposed to test, 40 files in tests

380 tests was the wrong ask. Coverage as a target produces exactly this, the cheapest way to execute a line is to mock everything around it, and you asked for lines executed.

Two changes fix it.

Ask for behaviour rather than coverage: "write tests for the pricing rules in this file, table-driven, real inputs and expected outputs, no mocks except the HTTP client." Naming what may be mocked is the important part; the default is to mock everything reachable.

Then verify with mutation testing rather than coverage. Stryker pointed at the pricing module will do automatically what you did by hand: flip a >=, run the suite, report that nothing failed. Start with one module, it's slow. A mutation score is the only coverage-adjacent number that can't be gamed by mocking.

Also: delete the 40 files. Seriously. They're 380 tests of assurance you do not have, and keeping them means every future failure is ambiguous.

The reliable pattern for me: write the first test myself, badly. One test, real inputs, the assertion I actually care about. Then hand it over with "add cases in this exact style, same structure, no new mocks."

Given a concrete example it copies the shape faithfully. Given a blank file and a coverage target it produces mock theatre. The failure is almost entirely about which of those two situations you put it in.

Make the bad pattern impossible rather than discouraged. A lint rule that bans vi.mock in files matching your rules/domain test glob means those tests physically cannot mock, so they have to compute.

Doesn't help with the existing 40 files, but it stops the next 40.

Your calculates tax correctly example is worth keeping as a teaching artefact. A test that mocks its subject and asserts on the mock is a tautology, it passes for every implementation, including no implementation at all. Not a weak test, a zero-information one.

The concrete tell: if you can delete the function body and the test still passes, delete the test.

84% coverage on an inherited service with zero tests, in one go, was always going to be false. Real coverage on legacy code needs refactoring for testability first, and the model routed around that by mocking - which is exactly what a person under time pressure does too.

coverage is a bad metric, everyone already knows this