Seoul cohort notes: why we cap concurrent labs

We run overlapping labs because engineers rarely work in isolation. The downside is noisy neighbors on shared service projects. Our answer is hard caps on concurrent mutations, not vague promises about infinite scale.

Instructors rotate through clusters hourly, snapshotting state so a single broken firewall rule does not cascade. Students see the activity log entries we use to narrate what happened, which builds operational instincts faster than polished slides.

We also publish teardown windows. If you know a classroom VPC disappears at midnight, you plan commits differently. That discipline carries into employer workstreams where change windows are real. Finally, we document when we refuse a request—usually when it would weaken another cohort isolation boundary. Saying no clearly is part of quality standards.