When backing up a csv-volume (in my case 3-node HA native-SAN) the process of generating the shadow copies consumes all memory on the addressed current node.
That leads to a node restart

That happens with any backup soft (Veeam, DPM ....) using the MS shadow copy providers.
MS solved that issue with the hotfix http://support.microsoft.com/kb/2813630 (in my case it solved that issue).
Ralf.