I did this for a while, but it masks subtle issues related to transaction boundaries. If your database usage is very straightforward, it may be adequate. But if

I did this for a while, but it masks subtle issues related to transaction boundaries. If your database usage is very straightforward, it may be adequate. But if you are, to name an example on the other end of the spectrum, in a monolith application that has pulled all the stops on fancy stuff like JPA callbacks, this should not be your approach. Also, don't rely on a local database, but rather on something that is reliably reproducible in your build, which will often make the discussed approach less attractive anyway. And if your database usage is really that straightforward, an in-memory database like H2 could serve one well.