Unfortunately I don’t have time to debug this further right now, but (in case anyone wants to try and reproduce this) I appear to have hit a problem like https://github.com/lxc/lxd/issues/7420 on an Ubuntu 20.04 server (Dell PowerEdge R210 II, lxd 4.0.3) and a single SATA SSD. After a kernel upgrade (but as far as I know, without any other configuration changes which may have triggered this).
The problem looks similar to this: https://github.com/lxc/lxd/issues/7420
All of the containers which had limits.disk.priority configured failed to start following a reboot.
New kernel package was linux-image-5.4.0-45-generic I’m not sure what the previous running kernel version was - may have been linux-image-5.4.0-42-generic or maybe linux-image-5.4.0-40-generic.
I can probably do some debugging on this some point next week if you think its useful, and this hasn’t shown up for anyone else by that time.