Comment 6 for bug 1928839

To sprinkle in a bit of cross-domain-responsibility-hatred on top of this suck-salad, it might not be a purely Curtin problem - restarting the MaaS host makes the behavior go away for a couple of iterations. I honestly see no rhyme or reason to that: its not like maas tells curtin to wait longer after maas is freshly booted.
After letting maas stay up for the last week, i started seeing this manifest on properly backed control VMs during provisioning (they live on top of 10k disks in a ZFS mirrored stripe with SLOG and ARC) - so while slow storage seems to trip this issue much more often, it apparently can occur with actually fast disks underneath.