Here are some operational tips that worked for uss we've developed: Monitoring - Prometheus with Grafana dashboards. Alerting - PagerDuty with intelligent routing. Documentation - GitBook for public docs. Training - pairing sessions. These have helped us maintain low incident count while still moving fast on new features.

Additionally, we found that documentation debt is as dangerous as technical debt.

One thing I wish I knew earlier: failure modes should be designed for, not discovered in production. Would have saved us a lot of time.