Why DataLoader instances must be request-scoped when solving N+1 in code-first NestJS GraphQL.
In a code-first NestJS GraphQL app, you add a DataLoader to batch/cache DB calls in a @ResolveField to fix the N+1 problem. What is the critical reason DataLoader instances must be created per-request (e.g. via a REQUEST-scoped provider or GraphQL context factory) rather than registered as a default singleton provider?