Did it include UI tests which are notoriously difficult, and if so how did it handle issues like timeouts and the async nature of UI?
Did it have rigid separation of concepts between unit vs integration tests etc, or more fluid?
Could you refactor internal code without changing tests — the holy grail.