Divide deployment targets between required and optional
My environment has servers that sit in reserve and are only started when needed. A deployment right now consists of bringing these up, waiting, and pushing the deploy button, but this doesn't always go smoothly.
Sometimes the Octopus service has a hiccup on boot, or AWS takes longer than usual. This gets reflected in releases that fail which leaves me with either running it multiple times and either explicitly including or excluding servers or stopping and starting an instance until it resolves.
Being able to communicate to Octopus that as long as a core group of servers is communicative, the release can be a success. I can then reconcile errant instances at a much more leisurely pace.