Glancing at the JDK issues, I can see that we are hardly first to run into this issue :-/
Anyway, does that mean that using Weld shutdown hooks on Windows may yield unexpected results? Should we mention/dosument this in the API?

Regarding rest skip/removal - the test is a sequence of three methods, I would suggest detecting the OS it is running on and skipping the problematic assertion only. Removing the test is not a good idea as it makes sense to use it on Linux and to avoid execution on windows, we would need a new annotation or profile. And that is IMHO not worth it for a single test, the rest seems to be executing fine (contrary to what Jenkins job above shows; running the tests locally, only this one failed).