Teaching async boundaries without drowning newcomers

We start async sections with a deliberately small API surface. Participants implement a worker that respects cooperative cancellation, then layer structured logging so support engineers can trace failures without reproducing laptops-on-desk theatrics.

The second week focuses on failure injection: we break DNS, throttle disks, and ask teams to narrate what they observe. The goal is vocabulary, not heroics—clear words for stalls, timeouts, and abandoned tasks.

In the third segment, we compare fire-and-forget pitfalls with explicit task graphs. We keep examples grounded in Japanese retail scheduling because concrete nouns anchor memory better than abstract diagrams alone.

Finally, we close with a bilingual glossary of review phrases mentors actually use. The point is repeatable kindness in code review, not a theatrical rescue story.