One process per scheduled job, and the store each writes. A job that reads a store as well as writing one is deriving rather than retrieving. 1.3, 1.10, 1.11 and 1.14 read no external source: they compute from what the retrieving jobs have already stored. Sources are not redrawn here; their flows enter from the edge, named, and each is defined against its entity on the context diagram. Every job drills to Data Pipeline for its schedule and script path. Per-job process specs do not exist yet.