Reporting queries are saturating an RDS primary in a Multi-AZ instance deployment. What actually helps?

Why this is the answer

In a Multi-AZ instance deployment the standby exists purely for failover and serves no traffic, so adding availability does nothing for a read bottleneck. A read replica is a separately addressable instance kept up to date asynchronously, and pointing reporting at it takes that load off the primary. Promoting the standby would break the failover pair, and backups are not a read path. (Note the contrast with a Multi-AZ DB cluster deployment, whose two reader instances are readable by design.)