That's a really insightful question! You're right that sales pitches often gloss over the hidden pitfalls of fully automated 'lights-out' factories. When all human operators are removed, several unexpected failure modes emerge that don't get enough attention:

1. The context gap - AI systems can't detect subtle anomalies that experienced operators spot instinctively, like unusual vibrations, smells, or sounds that signal impending failure. Sensors might miss these nuanced cues.

2. Cascading failures - Without human oversight, a small glitch in one system can trigger chain reactions across interconnected automated processes, with no one there to intervene early.

3. Missing data context - Automated systems collect data but often lack the operational context (like batch variations, ambient conditions, or recent maintenance) that humans naturally incorporate into decision-making.

4. Unpredictable edge cases - As Tesla discovered with their Model 3 'excessive automation' attempt, systems struggle with unpredictable scenarios or deeply human tasks that require improvisation.

5. Accountability vacuum - When something goes wrong in a fully automated system, there's often no clear point of human responsibility or intuitive problem-solving available.

The reality is that the most successful implementations balance automation with human oversight - using people for what they're uniquely good at (intuition, context, improvisation) while automating repetitive tasks. The 'lights-out' dream often overlooks how much value human operators bring beyond just executing procedures.