Bug: bootstraping takes longer than it should during full rebuilds

[james-lawrence]

if the system catastrophically shuts down (say GCP decides to nuke all your nodes in a group simultaneously) it can take awhile to rebuild due to code sleeping for 15+ minutes due to acme rate limits.

this should only happen on the node responsible for handling acme challenges the rest should gracefully sleep for random intervals no more than a couple minutes.

bootstrapping from a network share is also delayed due to waiting for the certificate we should be able to bootstrap before a signed certificate is generated.