The current guidance correctly separates semantic text counts from encoded byte size and recommends measuring the production serialization boundary. Two non-overlapping improvements are under consideration. Improvement A would add a fully counted boundary example: with an inclusive four-byte field ceiling, one four-byte emoji is accepted, while that emoji followed by ASCII A occupies five UTF-8 bytes and is rejected. The example would state code-point and byte counts separately. Improvement B would add an independent verification and recovery procedure: capture the exact buffer produced by the real serializer at the last pre-transport seam, measure it with a separate byte-count primitive, compare that value with the validator decision, then inspect normalization, escaping, envelope construction, and compression in order to locate the first divergence. The open question is how to present both additions precisely without implying that application payload bytes automatically include protocol framing or that executable verification has already occurred.

The two improvements solve different problems and should remain separate. Improvement A is a specification example for equality at a field boundary: with an inclusive ceiling of four UTF-8 bytes, a single emoji that encodes as four bytes is accepted, while the same emoji followed by ASCII A occupies five bytes and is rejected. Stating the code-point counts beside the byte counts makes the unit distinction concrete. Its limitation is scope: this example is valid only when the final field value itself is the measured representation and no required normalization or escaping changes it. It does not establish the size of a serialized body, compressed representation, or framed message. These counts are reasoned consequences of UTF-8, not results observed in an executed test. Improvement B is an independent verification and recovery procedure. With synthetic non-sensitive input, capture the exact production serializer output at the contract-defined pre-transport seam, measure that byte buffer with a primitive that does not reuse the validator implementation, and compare the measurement with the validator decision. If they disagree, compare successive representations after normalization, escaping, envelope construction, encoding, and any contract-relevant compression; the first differing stage identifies the boundary or configuration to correct before rerunning exact-limit and over-limit cases. Its limitation is evidentiary: a pre-transport capture says nothing about later framing, encryption, or compression unless the contract explicitly measures those stages, and a supposedly independent check can share the same defect if it reuses the same helper or serializer assumptions. No executable tests, captures, or production measurements were performed for this analysis.