# Tiered fee local validation

Reviewed 25 October 2026. All tariff/usage examples are synthetic Cedar inputs.

- 101 independently constructed complete expected results match Python and TypeScript, including 23 accepted calculations. Strict TypeScript compilation passes. Four Python behavioral suites pass.
- Integer/Fraction oracles predate implementation. Boundaries cover marginal/whole-amount changes, zero exemption, minimum/cap order, aggregate half-even ties, above-safe-integer arithmetic, blocked posting amount, dates, malformed tiers and exact keys.
- The actual compiled TypeScript reference embedded in the lab matches all 101 complete fixture results at each of 390, 768 and 1440 px (303 comparisons). Seven scenarios provide 153 guided states across those widths.
- Scenario/method changes stop playback; Back/Step/Play/Pause/Reset, keyboard and reduced-motion behavior pass. Controls meet 44 px minimum targets. Large amounts remain readable, the audit table scrolls locally, and there are no browser warnings/errors or document overflow.
- The static allocation comparison, rendered Mermaid workflow and phone/wide lab screenshots were inspected. The legend uses the actual number of tiers in the selected scenario.
- Package/playground validators and the isolated visitor build are separately recorded domain integration gates.

Raw receipts: research/test-results.json and research/browser-qa/results.json. Reproduce core checks with tests/run_tests.py; browser checks use qa_tiered_fee.mjs in the domain production report.

This is a pure calculation and uncommitted proposal. It does not prove tariff entitlement, complete usage, customer authorization, durable acceptance, or a real provider charge. No rendered video is claimed.