This makes slapos node instance slower, especially when there are many
kvm partitions on the node.

We don't really need to sleep because if qemu is not ready, promise will
fail and succeed on next run.