Containers: smaller images, calmer deploys

Large images are not just slow; they hide outdated libraries. We set byte budgets and enforce them in review, not as shame metrics but as shared hygiene. Health checks should reflect dependency readiness, not just “process up”. We walk through failing cases: database not reachable, migrations mid-flight, cache cold starts. Participants script probes that fail loudly in staging first. The cohort leaves with a runbook skeleton they can adapt — Malaysian SMEs rarely have dedicated SREs, so clarity beats clever automation.