Designing resumable large-data pipelines

This guide begins with a job stopped halfway through an import. It develops the state and restart rules needed to identify accepted work, isolate failed records and continue without rebuilding the entire run.