Plan and verify a data backfill — recomputing or reloading historical rows/partitions after a logic fix, new column, late data, or corrected source — safely, idempotently, and without corrupting good data or blowing the budget. Use when asked to "backfill", "recompute history", "restate the numbers", or reload a range after fixing a bug. Produces a backfill plan with scope, batching, idempotency proof, cost/load estimate, a validation strategy, and a rollback path. For designing the ongoing incremental pipeline itself, use dataform-bigquery-development or db-to-dwh-pipeline.
