We once hit a global database deadlock during a production deployment — everything froze, latency spiked, and services started waiting on locked transactions.
First step was calm response: stop traffic, notify the team, and pause the rollout to prevent further damage.
Then we traced the issue to a migration causing blocking queries. We rolled back to the last stable version, fixed the query design (added proper indexing and reduced lock scope), and redeployed safely.
Lesson: most deployment failures aren’t about bad systems — they’re about missing rollback plans, weak observability, and unoptimized migrations.
More context: https://go-cloud.io/