Scenario: You have developed a new version of a Dataflow streaming data ingestion pipeline that reads from Pub/Sub and writes to BigQuery. The previous version of the pipeline, which is currently running in production, uses a 5-minute window for processing. You need to deploy the new version of the pipeline while ensuring that no data is lost, there are no inconsistencies, and processing latency does not increase by more than 10 minutes. Question: What steps should you take to deploy the new pipeline version?