We use auto-deploy to make sure that new AWS auto-scaled machines get the current release. We however see strange results when a previous deployment has used guided failure mode, seemingly these auto-deployed machines enter a failure guidance mode when they have deployment issues which blocks any further auto deploy triggers and hence any auto scaling events. This seems to happen regardless of the failure guidance mode for the environment.
The scenario is that a careful developer has pushed a release in guided failure mode so that they can shepherd the release through if necessary. Some time later an auto-trigger deploy fails because some other reason (in our case it seemed like some octopus tentacle upgrade blocked a deploy somehow). Our auto-deploy failures are ultimately recoverable because we use ASG lifecycle hooks. All subsequent deploys including auto-deploys are blocked, those machines will get terminated automatically by the ASG.
Is this behaviour expected? I would think that releases from auto deploy triggers should never be in guided failure mode as the humans are in bed.