I reduced monitoring noise in the shared Apollo error handler by filtering `GRAPHQL_VALIDATION_FAILED` GraphQL errors from Sentry capture, alongside existing expected-error filters. These errors are typically frontend/query-shape validation failures that are non-actionable at runtime and can otherwise flood error monitoring.
I also added a corresponding ApolloFactory test case to lock this behavior and ensure `onError` still fires for UI handling while avoiding escalation paths intended for unexpected server failures.