Purpose-Bound Data Pipelines with Dataloop

Paper 6 of the ten-paper aZTF Research Series. Version 1.0-preprint. A temporary processing workflow can finish its computation while leaving its purpose, effects or evidence unresolved. This paper applies the author's aZTF framework to a synthetic engineering comparison and examines Dataloop as a proposed processing component. Dataloop developer tutorials describe graph composition and execution; separately identified DDOE documentation describes cycles typically tied to an item and their node executions while qualifying pause, changed-composition resumption and default accounting. Those limits motivate an inspectable contract connecting current task authority, the exact composition consumed, item-level paths, quality decisions, target effects and independently retained evidence. Physical hardware and network eligibility precede logical narrowing, and every model remains inside externally enforced rails. Reconstruction of a removed worker's history is distinct from rehydration under fresh eligibility and authority. The paper builds on existing workflow and provenance research and the series' previously proposed ephemeral-workflow lifecycle; it claims no invention of those foundations. Its evaluation separates finite conformance from a population comparison, gives a credible conventional engine equivalent enforcement and reconstruction opportunities, and ranks cost only after declared quality and reference-coverage conditions. The unit is a completed engineering task both accepted by its accountable engineer and supported by an independent reference assessment under the declared rule, with failed work, retries, human review and custody costs retained. No deployment, control test, reconstruction or comparative experiment is reported as executed. The result is a proposed processing and evidence contract whose utility remains open to selected-build testing and an unfavorable comparison.