Comment 9 for bug 1031065

Finding a better solution to the 'start networking' that we have here has lead me to open bug 1047712.
The case that I was debugging was actually not even improved by the 'start networking'. The issue that lead us here was that the ephemeral images for maas that we were trying to refresh have had issues coming up and woudl hit this timeout.

slangasek helped me to come up with a solution that has gotten us closer.
I've replaced 'start networking' with "virtual-filesystems-emitter" (see attached) in the cloud-init-nonet job.
Essentially that just gets virtual-filesystems event emitted if it hasnt already been emitted, which jump starts the networking code even while mounted MOUNTPOINT=/ is being processed.