You can still write unittest and launch them with pytest, you will get immediately:
much better output in case of failure
autodiscovery of test
By using the pytest the way it's meant to be used:
Simpler code
fixtures :
play several times the test with different data
scope : you can chose to build a object once in a play, once per file or for each file.
a lot of plugins : coverage analysis, mocks... inso as integrated some interesting things for sakia.
other things I don't know yet ;-)
I think we shouldn't wait to write tests : can I issue a Pull request you with just one or two test to build the test structure ? we can add some tests at each merge request after that.
Since it's on separate files we can merge the request at any time, I mean not like the black integration