In the context of designing a robust data pipeline that processes large volumes of data from multiple sources and writes the results to a Delta table, you are tasked with ensuring atomicity and consistency of data writes, especially in scenarios involving failures. The solution must not only guarantee that all changes are committed or none at all to prevent partial writes but also align with best practices for scalability and cost-efficiency. Considering these requirements, which of the following approaches best leverages Delta Lake's features to achieve atomic and consistent data writes? (Choose one option)