@sharwell I have used xunit v2 parallelization, it is definitely handy for getting long test runtimes down (not sure if this is a problem for you). We had some integration tests that would take ~35 minutes to run because each test would take 30-120 seconds to run but most of the time was spent waiting for external resources. We converted them to xUnit v2 and got runtime down to ~3.5 minutes with the CPU pegged the whole time.
@Zoltu Long tests are actually a problem in this project, but I think most of them are CPU intensive, so I'm not sure that we would be able gain any benefits from parallelization, but an ability to run only specific test case is very useful.
@SergeyTeplyakov The gains from CPU intensive tests are definitely limited. However, I have found that often times tests will only burn CPU on one thread in which case you can often get performance gains by running a number of tests equal to your core count. I believe xUnit does this by default (one parallelized test per core).
That all being said, I'm definitely a huge fan of being able to run one-off tests. It makes debugging and iterating on them so much easier.