AstralLight: Selected Supporting Evidence for Policy-Evolution Consistency and Fail-Closed Fault Containment

Selected supporting evidence for the AstralLight Journal of Systems Architecture manuscript, “AstralLight: A Host-Integrated Authorization Runtime for Policy-Evolution Consistency and Fail-Closed Fault Containment.” This dataset contains selected native-suite and restricted common-subset CSV files, fresh-state run-level summaries for RQ1–RQ4, ablation summaries, Redis process fault summaries, stage-breakdown data, data dictionaries, claim mappings, and protocol documentation. The evidence is bounded to the specified single-node host, synthetic workload, dated campaigns, configurations, JAR provenance where recorded, and adapter paths. The package supports inspection of the reported engineering observations but does not contain the complete host-integration implementation, complete Java benchmark reactor, deployment environment, compiled JARs, credentials, or all provenance inputs required for an exact rerun. Campaigns and operation-level statistics are kept separate and are not pooled across protocols. The common-subset comparison is a restricted single-run adapter-path comparison rather than a universal performance ranking. The Redis-down phase contains no decision-operation samples and therefore does not provide request-level outage latency. The 99.9% and 97.4% reductions reported in the manuscript are analytical logical-row counts, not database-byte measurements. No source code or executable benchmark implementation is included in this dataset.