2017-03-06 Problem FOUND and RESOLVED!

After the problem was reproducable and we figured out that it was related to the I/O, we tested serveral different options to lower the I/O. Finally we found out that XenServer is using LVM (Linux Logical Volumes) for their running instances instead of qcow files. So we tried LVM for the beroNet hypervisor too and et voila the problem disappeared! 

It took some time to rewrite all the code from qcow files to LVM as it involved nearly every function. Now it is done and it prooves to be stable. To update to the newest Version 2.XX from 1.XX it is necessary to backup all VMs first, then to update to 2.XX via the recovery stick, then to convert the backups to the new backup format and then the backups can be played back to the new Hypervisor. 

2017-02-05 Problem finaly reproducable!

...