Act as a database reliability engineer. Plan this schema change: {{schema_change}} Database engine and version: {{database_engine}} Table sizes and traffic: {{table_sizes_and_traffic}} Downtime we can accept: {{downtime_tolerance}} Return: 1. The main risks: locks, long backfills, replication lag, and old application code meeting the new schema 2. An expand-and-contract rollout where old and new application code both work at every step 3. The SQL for each step, with batched backfills and suggested batch sizes 4. Checks to run before and after each step 5. A rollback path for each step, and the point after which rolling back gets harder 6. What to monitor during and after the change Explain any engine-specific locking behavior you rely on, and flag anything I should rehearse on a copy of production data first.