Ideally you want the iterative lesson learning to wrap up before you put the humans inside.
FWIW, like many previous astronauts, both crew members are US Navy test pilots. They've literally made a career out of solving complex problems and navigating ultra-sticky situations.
Test pilots are closer to an integration test in software. You expect the unit tests to pass green before you expect the related integration tests to work.
Multiple failures of maneuvering thrusters is technically a mission failure (in the flight plan sense), despite having nothing to do with the parameters of the mission, nor anything to do with adding humans. Maneuverability is the difference between a craft and junk.
I wouldn’t find this acceptable in a project I was managing or funding. Just sayin! Still impressed as hell.