Part of the issues that we seem to be running into with the amulet testing is that Juju returns complete before a deployment is actually complete. This is by design (run 'juju deploy <someservice>' and see how quickly it returns). Because of this, amulet has no way of knowing when a deployment is actually complete and when to start running tests. After speaking to some of the Juju/amulet devs, it is my understanding that this is what sentries are for in amulet. I tried switching from sentries=False to sentries=True, however things blew up, so as a quick work around per some of the Juju devs, I just implemented a wait for now. We really should investigate switching to using sentries as they are designed to work for this purpose.