After debugging for more then a week I finally found the problem, well actually multiple problems.
The main problem is that the Aruba VAs use an ancient linux kernel that doesn't suppor the default settings of modern KVM hypervisors, namely the Q35 chipset. The main difference beeing that Q35 uses PCI-E whereas the old I440FX uses only PCI.
This in turn causes the 2nd error, namely that the old kernel can't access any virtio devices that with the new chipset all use PCI-E which the old kernel can't access. This leads to the problem that the massive and huge rcS init script can't find any disks.
But the rcS init scrip has a flaw, it doesn't verify the output of "fdisk -l" which causes the 3rd error. The rcS script silently gets stuck in a endless loop calling "sleep 1" as it tries to wait for a non existend device "/dev/a1" to appear.
To add a bit of context, The 3.18.26 Linux Kernel that is used in the MM-VA is end-of-life since February 8, 2017.