Queues are a teaching tool, not a performance flex

Queues appear late in our syllabus because they amplify whatever discipline already exists. If logging is messy, queued jobs only hide failures for longer. We assign small failure-injection exercises so retry counts become visible facts instead of myths.

In the workshop week we pair Horizon-style dashboards with a simple spreadsheet of alert thresholds. Participants decide what should page humans versus what should self-heal, then defend those choices to a mentor playing the role of a skeptical product manager.

We close with a migration path: how to move a cron-shaped script into a job without changing business outcomes on day one. That operational framing keeps the conversation away from buzzwords and anchored in supportable systems.