Which strategy optimizes memory consumption during GraphQL schema generation for large datasets in Gatsby v5?

Gatsby uses a centralized GraphQL schema to unify data from various sources during the build process. When utilizing the sourceNode API to inject extensive custom datasets into the internal store, the memory overhead during the schema generation phase can increase significantly.

While the .cache folder optimizes incremental builds, the initial processing of large-scale data sources often pushes Node.js heap limits, necessitating the use of --max-old-space-size. There is a need to determine if there are architectural patterns within the Node API to reduce this memory footprint without sacrificing the ability to perform complex queries across the entire dataset.

• Does Gatsby provide a mechanism to stream data into the GraphQL layer to avoid loading entire datasets into memory?
• What are the trade-offs between using createPages with large data arrays versus utilizing a more granular sourcing approach?