How do you handle database normalization versus denormalization in a high-traffic environment for Chargebee's billing and subscription data?
