It seems that during test execution of the core module (master branch) the number of used threads ramps up quite noticeably.
On Linux this is likely to hit the barrier of permitted threads per process, which will prevent the GC thread to start and in turn send the testsuite to critical failure via a very misleading OOM exception.

you can observe the behaviour running the testsuite and attaching jconsole to it, on the "threads" tab you'll see the number raising. In my case it goes up to around ~800 threads before being killed (long before the tests are finished).