On slow boots snapd falls into a initialization loop

Hi,
while debugging other issues I had a guest slowed down, so the setup would be like:

1. KVM guest with Cosmic (4 CPUs 2G memory)
2. slow down this guest by ~1/8 with cgroups (see https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1732028/comments/13)
3. inside this KVM guest run a KVM guest with TCG (no -enable-kvm) with Cosmic as well

What happens is that on early boot things are fine but when snapd is supposed to initialize it fails and enters a init-loop trying over and over again preventing the system to reach a usable state to log in.

Looks like this:
[FAILED] Failed to start Snappy daemon.
See 'systemctl status snapd.service' for details.
[ OK ] Stopped Snappy daemon.
         Starting Snappy daemon...
[ OK ] Started OpenBSD Secure Shell server.
[FAILED] Failed to start Snappy daemon.
See 'systemctl status snapd.service' for details.
[ OK ] Stopped Snappy daemon.
         Starting Snappy daemon...
[FAILED] Failed to start Snappy daemon.
See 'systemctl status snapd.service' for details.
[...]