AIAP: Cause AIAP to fail faster

[ian-howell]

Problem description (if applicable)
Currently, when a container within AIAP runs into an error, it continues to silently run, without actually doing anything (via tail-ing /dev/null). This creates a problem when watching for failures in a deployment (e.g. gating), since there's no way to know whether the deployment has failed or not without operator intervention.

Proposed change
I would like for AIAP to fail as soon as one of the containers fails.