So this seems like a good idea, but I notice that all tests are marked
as failed until the results arrive. This leads to incorrect failure
emails and incorrect UI indicating lots of test failures where none
exist. Doesn't seem ready for production yet.