Problem
Paradox Shop's nightly bulk import worker writes at 8,000 rows/second to the primary, while the storefront now serves 500 requests/second of customer-facing reads from a Small read replica. The app tier was sized for launch-day traffic, the primary itself can't absorb 8,000 writes/second directly, and the replica's apply capacity is badly outrun by the import on top of that, leaving customer-facing reads minutes stale until it recovers. Resize the storefront's fleet, the primary, and the replica so this actually holds up.