The tl;dr is that "LXC defaults to providing 4 ttys, LXD on the other hand doesn't provide any". This means that the containers that have been migrated from LXC to LXD have services defined for 4 getty processes, but there's no backing device to run them so the procs just spin causing excessive load on the host, and vast amounts of logs filling up the disk.
We can get over this by removing the tty* services from /etc/init in the container (and stopping the services if needs be).
See https://github.com/lxc/lxc/issues/1096 for some illustration here.
The tl;dr is that "LXC defaults to providing 4 ttys, LXD on the other hand doesn't provide any". This means that the containers that have been migrated from LXC to LXD have services defined for 4 getty processes, but there's no backing device to run them so the procs just spin causing excessive load on the host, and vast amounts of logs filling up the disk.
We can get over this by removing the tty* services from /etc/init in the container (and stopping the services if needs be).