How do unit, integration and end-to-end tests differ?

A unit test checks a small piece of logic, e.g. calculating a limit. An integration test checks how parts work together, e.g. an API with a real test database. An end-to-end test walks through an important user journey, e.g. logging in, submitting a form and seeing the result.

Each answers a different question. A fast function test won't prove the app connects to the database correctly. On the other hand, a single long browser test can make it hard to find the cause of a failure. I match the test level to the risk and check specific behaviour, not just that the program “doesn't crash”.