A high-volume application generates excessive telemetry. The team wants to control ingestion while retaining a useful view of distributed operations. Which approach should it evaluate?

1. Delete all performance metrics permanently
2. Log every secret to compensate for sampling
3. Appropriate telemetry sampling with attention to correlation and analytical needs
4. Stop recording all errors without evaluation