> Wonder if the wrapper should have a mode where it retries the test once when
> it failed, this way we don't incur any overhead for tests until they fail
> (obviously could miss recording some failures, but maybe that's ok).
This is going to be hard using that kind of wrapper. What we could do is to first run all xvfb* targets and then running the rxvfb* of the failing tests.
> Wonder if the wrapper should have a mode where it retries the test once when
> it failed, this way we don't incur any overhead for tests until they fail
> (obviously could miss recording some failures, but maybe that's ok).
This is going to be hard using that kind of wrapper. What we could do is to first run all xvfb* targets and then running the rxvfb* of the failing tests.