Reproducing test in https://github.com/LittleJohnII/wildfly/tree/eap7-495
The watermark test will fail and logs will show that work wasn't distributed to the second node when the amount of threads available got down to 1 (and watermark = 1). Even with the default watermark setting, the work is never distributed - if thread pool size = 2 and 3 instances of work are executed, the third always waits for the previous two to finish on the first node.
Stefano Maestri is already aware of the issue - it was discussed on HipChat prior to the Jira creation, so this is mainly a tracker.
I'm setting blocker priority since I'd like to have watermark working before EAP7-495 is resolved (I think this is the most important policy setting, and the default too). If this is a problem, we can discuss lowering the priority and pushing EAP7-495 out anyway, since this isn't a thing that breaks the whole DWM.