Evidence and verification
This is where I keep the questions behind the Library. Each one is marked as completed, in progress, or waiting, with the evidence still needed to settle it.
Every later formula and mod claim depends on the correct current baseline.
Book XIII supplies the complete official patch side; Book XIV reconciles every patch token against the current auxiliary manuals.
A reversed order can change hundreds of shared objects.
Project description and version 1.15.0 changelog, audited in the independent integrity report.
The collision breaks the identity shared by DE's Ring, Amulet, and Blood Vial helpers and the Divinitus Crystal Priest.
Exact source trace, upstream accepted repair, and pinned Inspector resolution in the independent integrity report.
Stable website records and downstream references require versioned secondary identities.
Exact-version structured Inspector register in the independent integrity report; engine confirmation remains separately labelled pending.
An unfamiliar command cannot be explained safely until documentation, parser recognition, annotation syntax, and likely errors are separated.
Static command inventory against current official manuals, patch notes, and the pinned parser in the independent integrity report.
Pretender and fort economics depend on the correct scale multiplier.
Modding Manual 6.34 defines `#deathincome` default 2; no official announcement through 6.36 changes it.
Production planning and scale evaluation depend on the current value.
Revision-2 main-manual 2% rule checked against the complete official patch ledger through 6.36.
Close recruitment and fort calculations can change at integer boundaries.
The canonical section now separates six rounding subsystems and their operational displays; complete closure still requires published 6.36 boundary reproductions or developer-confirmed formulas.
Commander bottlenecks can determine fort and mage efficiency.
Revision-2 fort bonuses, current 6.35 community base and accumulation reference, official `#slowrec` and `#rpcost` semantics, AI exclusions, and patch review through 6.36.
Persistent roster costs affect long-war sustainability.
The 6.33 observation and pinned 6.35 component fields resolve stacking and mounted decomposition; annual samples exclude a simple final floor, while shape classes and monthly aggregation remain open.
A province can appear affordable while being unable to recruit the planned roster.
Official shutdown and current community reduction claims are preserved; the loss-floor reading now has explicit 1-, 2-, and 3-point candidate thresholds, still awaiting a versioned reproduction.
Supply range determines operational reach and siege endurance.
Range, maximum distance, and highest-fort selection are separated from the official `x6` versus `x4` multiplier conflict; a published 6.36 distance series remains required.
Commander survival affects retreats, sieges, and remote expeditions.
Need Not Eat, the 6.18 transformation correction, 6.35 object fields, and the 6.35 equipment-refresh fix resolve explicit immunity and display state; generic priority remains open.
Opening timing determines whether a design solves expansion or only a later war.
The June 2026 Dominions 6 page republishes Loggy's model, but source tracing reaches pre-Dominions 6 notes, no 6.x raw sample, and a one-turn range discrepancy. Current reproduction or developer confirmation is still required.
Temple, priest, and candle investments depend on the actual spread model.
Revision-2 official formulas and turn sequence, reconciled with the 6.13 random-preaching correction and later patch history.
Small errors can invalidate a Pretender design or sacred package.
6.35 structured bless data plus official rules and patches through 6.36 establish activation gates, automatic blessing, repeatability, overlap, weapon-trigger, item, and mount boundaries; adjacent transformation and cross-source stacking remain separate questions.
Recovery planning after a god's death depends on time-to-return.
Official Holy, Disciple-game, Elegist, Recall God, Trinity, and hosting rules are reconciled with published controlled research establishing a fixed 50-point base and per-priest effective-rating `-1/0/+1` variation; detailed routing and placement remain Current Community Reference.
The result can decide a game despite later movement or siege outcomes.
Official steps 8 and 57, full and partial ownership rules, the current conquest rule, and a published same-turn report establish the claimant-death, unfortified-conquest, siege, and fort-storm branches. Exact simultaneous winning-total ties remain separate.
Rout thresholds influence squad sizing, standards, and casualty tolerance.
Current community formula supplies `round(5 x surviving proportion)`; a current raw boundary set and separate commander treatment are still required.
Fatigue breakpoints shape armour, melee endurance, and spell-support value.
Current community reference supplies one per active round plus Reinvigoration; a frame-by-frame 6.36 reproduction remains required.
Defence collapse and swarm counters depend on more than the visible attack value.
Official accumulation is resolved; a Dominions 5 decay constant is retained only as a regression hypothesis for a current instrumented reproduction.
Attack rate changes damage delivery, repel exposure, and fatigue pressure.
Current structured data and published timing reproduction.
Rider and mount statistics otherwise produce misleading army comparisons.
Official targeting and component rules are resolved; carry, mixed-component movement, and exceptional modifier order remain open.
A scripted spell package can fail because the caster conserves a gem.
Official 6.12, 6.15, and 6.30 corrections bound the historical thresholds; a current reproducible battle set remains required.
Boundary values can change caster safety and access ladders.
Published current reproductions separated by subsystem.
Communion failure can destroy the most valuable mage concentration in an army.
Manual rules and official 6.12-6.30 corrections resolve the core; automatic-item, form-change, and forced-removal behaviour remain open.
Movement prediction determines reinforcement and retreat planning.
A published chassis-value algorithm supplies a precise hypothesis; a 6.36 map-and-save reproduction remains required.
Counter-raiding and scout survival depend on risk rather than a binary threshold.
The official open-ended 2d25 formula has been enumerated directly from D=-60 to +60; exceptional contributors remain regression tests.
Economic warfare cannot be compared honestly without expected return and population damage.
Official legality and sequence are resolved; historical exact-output formulas are now 6.36 reproduction predicates rather than current law.
Interface-enforced treaties must not be described through informal multiplayer custom.
Official 6.01-6.30 exceptions are versioned; the complete binding, nonbinding, and absent action matrix remains open.
Endgame plans require exact Ascension Points, defenders, and claimed bonuses.
Pinned 6.35 structured site data, cross-checked against current official rules.
Remote warfare portfolios depend on cumulative interception risk.
Published current controlled reproduction.
Late-game plans can waste rare Astral access on obsolete or ambiguous requests.
Current official data or versioned published catalogue with reproduction.
Multiplayer races need a defensible risk model.
Official hosting rule or published simultaneous-order reproduction.
Commander survival packages depend on more than nominal bodyguard count.
Published current controlled reproduction.
Endgame clocks and coalition decisions depend on the actual failure state.
Current official object data and patch notes.
Large overhaul compatibility can fail without any numeric collision.
Author documentation or published deterministic harness.
Name references can fail even when numeric IDs do not collide.
Official parser documentation or minimal published loader reproduction.
A visible final card may omit inherited behaviour needed for compatibility.
Exact-version source history plus structured engine export.
Shape chains can silently lose national or sacred roles.
Published current transformation matrix.
Long event campaigns can corrupt state near numeric boundaries.
Official limit documentation or minimal published harness.
Scenario design needs a practical performance budget.
Published hardware, turn, province, and event-count benchmark.
Scenario claims should distinguish configured hints from observed behaviour.
Published behavioural sample tied to version and map.
Release updates need a safe migration policy.
Published version-to-version save matrix.
Public mod releases need predictable multiplayer installation behaviour.
Official documentation or published clean-client matrix.
A public veteran calculator must not conceal a version conflict.
Official 6.13 changed veteran HP to +1 at each star from the third onward, 6.14 corrected XP HP-stat handling, and the current manual repeats the three increments.
Old reverse engineering is useful but cannot be represented as current official law.
Edition 28 defines the official boundary and a versioned raw-observation template; current reproducible tests or engine-backed reverse engineering must supply any formula.
A false equivalence can invalidate targeting and counter design.
Official 6.36 ability definitions, current spell descriptions, and a patch check establish the category division; two narrower runtime edges moved to R-058.
Duplicate investment can waste gems, design points, and scripts.
Current definitions confirm best-only Standard and Inspiring Researcher, cumulative experience and Horror Mark, and non-stacking Blur, Displacement, and Invisibility; unlisted abilities remain test-specific.
Veteran timing and Hall-of-Fame plans depend on the actual gain sources.
Controlled monthly observation and current data export.
Visual or taxonomic assumptions can produce false counters.
Current object export plus controlled spell and condition matrix.
Mounted commanders can change body and function across the turn boundary.
Controlled mounted-commander series under 6.36.
Expansion, fort placement, and off-path recruitment depend on exact independent access.
Exact-version engine or inspector export containing population-type membership.
A spell record is incomplete when it names a selector but not the possible summoned units.
Exact-version engine export or published reproducible selector resolution.
Public filters should be readable without pretending that an inferred label is official.
Current UI wording, official manuals, and exact-version object export compared field by field.
Source completeness is settled, but accurate filters improve website maintenance and command chronology.
Human review of each official source bullet against the Book XIII classification key and canonical ownership map.
Search is complete, but a token's official existence is not always a complete behavioural contract.
Later official manuals or announcements, developer clarification, or a published minimal reproduction tied to an exact version and object context.
Category boundaries are documented, but these exact interactions still affect counter selection and strategic detection.
The manual explicitly confirms Arcoscephale dominion scrying and The Eyes of God as narrow channels; reproduce those controls and run the Edition 29 battle and information matrix for the remaining edges.
