Maintaining evidence dependent comparisons of domain registration prices

An advertised registration price and a renewal price answer different cost questions. A comparison computed from both can also outlive the evidence on which it depends. This paper studies these two failure modes using a small comparison engine that combines interval reasoning with versioned field dependencies. Its output is a relation between conditional posted-rate projections, not a guarantee about checkout costs or future prices. An exploratory acquisition from two registrar websites yields 1,028 observations for 514 common domain suffixes. Exhaustive masking of renewal fields across three horizons produces 6,168 scenario-mask requests. Carrying registration prices forward into missing renewal fields disagrees with the complete-snapshot reference on 965 requests. Interval comparison and a horizon-aware complete-case method both issue 3,084 conclusions, with no disagreement against that reference; the interval rule provides no coverage improvement in this corpus. In a controlled expiry-and-restoration replay, dependency-based recomputation agrees with full recomputation after all 2,056 events while reevaluating 4,112 rather than 3,170,352 query instances. These are operation counts, not measured speedups. The contribution is a reproducible application study that separates numerical ordering, evidential support, and claim maintenance. Its results support explicit withdrawal of expired claims but do not establish market-wide accuracy, improved consumer decisions, or a new general algorithm for uncertain ranking. This preprint is accompanied by frozen numerical observations, source identifiers, a documented protocol, Python code, tests and per-request results. Controlled masking and expiry events are simulated; they are not historical price changes. The manuscript has not undergone peer review.