A large enterprise org has a single corporate account with 50,000 child opportunity records. What architectural techniques should be recommended to mitigate data skew and sharing calculation delays?