Data and reproducibility code for "Release granularity and audit cost in aviation maintenance question answering"

This record provides the recorded research data, analysis code and supplementary methods for the manuscript "Release granularity and audit cost in aviation maintenance question answering". The deposit contains three archived data packages and four supplementary methods documents. Supplemental Data S2 carries the main validation: 24 source clusters, two local model deployments, three evidence conditions, five output policies, 432 recorded model calls, 720 policy versions and two independent coded rating layers. Supplemental Data S3 carries the frozen protocol extract, the field-level export map and three archived trace cases with a consistency checker. Supplemental Data S1 carries an earlier 24-task experiment that is reported as separately analysed supporting evidence; its source samples, executions and review layers are not pooled with S2. Version 1.0.1 updates documentation only. README.md was revised to correct the dataset contact email and to record the version history, and Supplemental Methods S1–S3 now use one consistent term for the same historical record state. Supplemental Methods S4 and the three data archives are byte-identical to version 1.0.0; no data, model call, rating, analysis or result has changed. Original research data and documentation are released under CC BY 4.0; original project code under the MIT License. Third-party source text, model templates and dependencies retain their upstream terms. Private consent correspondence, identity mappings and identifiable process receipts are excluded from this deposit. Offline reproduction requires Python 3.10 or newer and makes no network requests and no new model calls. See README.md for the current file, figure and table mappings.