At some point, every growing system hits the "you can't join that anymore" wall. Reads get offloaded to replicas, writes start queueing, and suddenly you're managing backpressure across half a dozen services.

Most teams respond by duct-taping workflows together with message queues, cron jobs, or hand-rolled orchestration logic. It's wild how quickly database scale issues turn into distributed state problems and how rarely people plan for that up front.

I've been leaning into explicit state machines lately (prop for making that less painful). Not because it’s trendy, but because once your system has implicit state transitions buried across services, observability and recovery get brutal.