Hi folks,
We have still been investigating this issue affecting a small number of Machines with an unavailable rootfs (/dev/vda) on a few hosts. It does seem that the previous manual fix we applied last week was reverted on the at least the ams host. I applied another fix which should once again unblock those instances (a few dozen total).
We’re hopeful the fix will be permanent this time, but if you continue to have trouble, you should be able to destroy the misbehaving Machine and launch a new one as well.