> Were you ever able to reliably reproduce the situation or fix it (other > than wait for many hours or installing a newer version)? NO. And we have more than one 400+ node clusters with week-long jobs, so believe me, we tried workarounds. Harald.