What is a Workflow State Machine?

Instead of representing a workflow as an imperative script that runs top to bottom, a state machine defines states like `pending_review`, `approved`, and `rejected`, along with the events that can move a workflow from one state to another. This makes illegal transitions structurally impossible: you cannot jump straight from `pending_review` to `paid` without passing through `approved`, because no such transition is defined. It also makes the current status of any in-flight process trivially queryable, since the state is just a field on a record rather than something inferred from log lines. State machines are especially valuable for long-running or asynchronous workflows that might pause for hours or days waiting on a human approval or an external webhook. Because the state is persisted rather than held in memory, the process can survive a server restart and pick back up exactly where it left off once the next event arrives. Well-designed state machines also define explicit error and timeout states, so a stuck workflow surfaces as a visible status rather than disappearing silently.