The vmTestbase/nsk/jvmti/GetThreadState/thrstat002 test is intermittently failing with the wrong thread state reported. It is most likely because of the unexpected JFR (or NMT) activity.
This was suggested by David H. :
  We need to dump the stack of the target thread when we find it in the wrong state. It is possible (in the absence of any detailed knowledge about how JFR might operate here) that JFR does cause the thread to go to unexpected code where it can be seen in the wrong state. But we really need to confirm that.