Declarative automation puts the "when should this update?" rule on the asset itself: @dg.asset(automation_condition=dg.AutomationCondition.eager()). A built-in sensor, default_automation_condition_sensor, evaluates every asset with a condition on each tick and requests the ones whose condition is true.
eager() requests an asset when any dependency has been updated (or will be requested in the same tick) and none is missing. Because a requested parent counts as an update, a whole chain is requested together, in one run, as soon as its root changes. It does not request an asset again while a request is already in flight, and on the very first evaluation it only records the current state.
In this workspace the clock moves only when you move it: Evaluate automation runs one tick, and runs complete before the next tick. The tick list shows, for each asset, which parts of its condition were true. A real deployment ticks about every 30 seconds.
When raw_orders is materialized, the next tick requests paid_orders and revenue in one run.