A financial services company runs a Cloud Bigtable instance that stores real-time transaction records. During a deployment, a faulty application bug corrupts a large set of rows over a two-hour window before the team detects the issue and halts the pipeline. Business requirements mandate that the data be recoverable to a consistent state prior to the corruption, with a recovery point objective of no more than a few hours. Which disaster recovery approach should the team have in place to meet this requirement most reliably?