Instead of running a full rebuild which can take a very long time when things are broken, use the last build. Capistrano stores the builds and it completely defeats the purpose of their existence if we never use them. It would take a few seconds at most to rollback a problem vs waiting for an entire new build. For our stack that's over 10 minutes of downtime.

My proposed solution based on some chats in Slack is to have both an Instant and Full Rollback option since I was told some stacks have deploy hooks that need to be rerun. A more advanced solution would be to give deploy hooks a configuration option to run on rollbacks.