I have taken the dual approach as !421 (closed) : rather than porting the existing hspec tests to tasty (i.e. embedding them via the tasty-hspec adaptor) I have done the opposite, i.e. I have tried porting everything to use hspec, discarding tasty altogether. Why?

given our previous concerns regarding the execution order of things, I've figured that hspec gave us more flexibility, because we can still keep the old sequential and parallel combinators, and adjusting them as we see fit, whereas tasty doesn't have that, and unfortunately the sequential and parallel combinators are lost when we use tasty-hspec which was the thing that kickstarted the use of two different drivers from the beginning.