To a degree. Certainly the end-to-end tests should be something you can expect to reuse. Other tests may be hit or miss. In the case of SQLite, where there's >100x as much test code as application code, there's a large range between "can use some of it" and "can use most of it" - as mathw gets at I think we should be able to expect the latter here.