27 September 2026 · Scenario and preserved method map

Model checks

  • Crossing composition: independent flat enumeration passed 14,262 queries and all 239 compatible witnesses. The numerical solver is unchanged.
  • Staged paths: independent all-history comparison passed 507 configurations, 7,405 plan replays and 38,308 stage annotations. All seven preset traces match the preceding edition after removing display labels; the current engine hash is recorded in its method note.
  • Programme studio: independent evaluator passed 29 queries, 19,941 candidate comparisons, 79,764 resource coordinates, 5,625 typed joins and 23,640 surviving monotonicity witnesses. Default: 504 candidates / 12 minimal alternatives. Baseline and unsupported-feedback counterexamples remain checked.
  • Monitoring power: 784 fresh calculations through the pinned Python package over 98 briefs; deterministic atlas regeneration passed. Independent enumeration checked 7,063,056 triples, 24,076 ceiling boundaries and every architecture record. Direct module-boundary checks passed 48 calls. The separate inert-DOM regression passed all 98 UI data projections, exact decimal ceilings, unit/provenance records and migration isolation.
  • Directory, specialist migration and co-design route checks passed locally. The site workflow runs these checks and the full Jekyll build again on the proposed revision.

The old optional browser scripts were updated but were not executed. Historical railway browser receipts do not establish a pass for the wildlife interface.

Fresh browser review

The current pages were inspected and operated through the visual browser against a local static preview, at the ordinary desktop viewport and a 390 × 844 mobile viewport. This was a bounded integration review, not an exhaustive accessibility audit or human-use study.

  • Home: emblem and scene loaded, the four distinct questions remained visible, and navigation reached the apps. Mobile page width stayed within the viewport.
  • Staged paths: advancing the default witness showed the guide-fence upgrade and active mobile monitoring team; the deadline counterexample reported “The endpoint fits. The journey does not.” The unavailable-team preset and conceptual explanation were also checked. The mobile diagram and explanation were legible without page-wide overflow.
  • Programme: default 504/12 result, actual component join table and both programme clocks were inspected. Lowering the target to 80 produced the sole (0,0,0,0) incremental-resource baseline. Reset restored the default. Method and Loop views responded to keyboard activation; mobile layout stayed within the viewport.
  • Monitoring: default £3,245 / 274 L; a 160 L ceiling selected £4,360 / 152 L. Zero demand produced £0; the long-deployment preset produced an explicit unavailable-catalogue outcome. The method view showed the energy/output/thermal inequalities; mobile layout and keyboard activation were checked.
  • Crossing: mixed-bundle default and source/method signposts inspected; target 330 produced explicit catalogue infeasibility (maximum 315 at three sites); target 80 restored the trade-off case. Method-tab keyboard activation and mobile layout checked.
  • Old rail-power address with old query keys remained a static notice. Following its replacement link opened the default monitoring brief with the new powerW query and no transferred footprint ceiling.
  • Browser error-log checks returned no errors. Temporary viewport override was reset.

Independent conceptual review

Separate reviewers checked the wildlife mappings, typed joins, limits and retained explanations. Two wording issues were corrected: “passage constraints” became construction-period readiness, and an inaccurate “habitat-footprint loop” description became the actual battery/conversion/cooling loop. No remaining material findings were reported in the reviewed HTML and shared method account.

The tests establish the declared finite calculations and observed interface behaviour. They do not establish ecological response, equipment suitability, stakeholder consent or practitioner adoption. Publication is checked separately after the repository’s normal checks and merge process.