The strict decoding error introduced in #587 is actually a permanent error. Even if the kubelet doesn't understand that yet, we should return that to the kubelet already.
This relates to the discussion we had in #375, in particular
We can slowly add them in preparation for the kubelet even supporting them at all.
Having the notion in our driver is there to future proof us once the kubelet does support this / not keep us looping in the driver unnecessarily on a permanent error