Hold The Horde — Campaign Evidence and State Crosswalk

Status. A finite index over content that already exists: it maps the eight accepted chains and fifteen primary cast scenes onto the actual rooms, controls, objects and observations they consume, and records where those interfaces agree or conflict. It is not a campaign map, a scheduler, new content, story truth, language grammar or a validation of anything. No new chain, person, room, mechanism, scene or object is created here. All five owner questions remain unanswered, and only their actual consumers are marked conditional.

Sources read live for this revision, with full hashes.

The index uses the source revisions above. A later change to a cited observation or state boundary requires checking its consumers; a new revision alone does not rename the underlying room, person or object.

Counts. 8 chains (L1–L4, S1–S4) · 15 primary scenes (Y1–Y3, A1–A3, D1–D3, M1–M3, I1–I3) · 26 evidence index entries, reusing their actual records and objects · 7 mutually exclusive rescue host variants · 5 Story30 personal variants · 2 dependency orders. Existing inventories are unchanged: 24 modules / 40 rooms, 10 encounters including 2 setpieces, 13 major mechanisms.

0. Identifier provenance

Established identities: scene IDs Y1–I3 belong to Cast and Short Rescue Map; R01–R40 and M01–M24 to Station Room and Module Inventory; objective numbers 1–12 to Regional Discovery Map. L1–L4 and S1–S4 are index labels here for the four numbered chains in Two Language Discovery Map and Story and Ending Map respectively. Those parents do not literally define these prefixed IDs. Room32 already uses L2 as shorthand for the second language chain; this index preserves that reference without inventing a parent-owned ID register.

Evidence aliases: every A-… key below is a locally created index handle for observations in the cited source descriptions. The source-owner table below names those descriptions and the player decisions they inform. A handle can group the readings needed for one comparison; it is not a new object or a promise that the game captures a whole sequence in one action. No established identity is renumbered, and parents need not adopt these aliases.

1. Chain and scene identity map

Scenes carry chain positions many-to-many. Story30 does not allocate its personal variants exclusively to one chain each, and reading them as 1:1 was an error in the earlier draft:

Host variants (7). R13 hosts Y1 or A1, never both; the other person takes R19 pressure annex (Yeva) or R22 supply booth (Aster). Abort hosts are the four already counted rooms — R06, R12, R24, R27 — and no scene adds a fifth.

2. Deduplicated evidence and knowledge ledger

One entry per observation; a shared observation is listed once with its separate consumers. All keys are local aliases (§0).

One observation, several kinds of consumer. A-TRANSFER feeds L3 as a role distinction and S3 as movement evidence; A-REPLY feeds L4 as register evidence and S2/S4 as evidence about what the line is; A-PROFILE feeds S2 causally and Y3/M3 as something a person reacts to. Shared observations are reused; distinct stages still require their own actual captures.

The chronology is three records, not a reader. A-STRIP, A-CARD and A-BERTH share explicit cross-references and together support old seal state → human profile installation → current docking registration for these records only. The player compares them in the paused journal; nothing declares the chronology true, and no diagnostic reads history.

Each comparison needs its actual observations. The initial side-by-side active-profile/bypass test needs both committed outcomes; it has no additional earlier-capture prerequisite. Comparing behavior before and after isolation does require a before-capture the player actually made. Two post-isolation results establish only current disconnection. No prior record is generated retroactively.

Source owners and decisions

These links own the underlying descriptions; the aliases remain local to this index. A record referenced by multiple rows is still the same record, and additional readings are not additional authored objects.

3. Representative state and commit boundaries

Three separations this crosswalk insists on. Holding an inference is not informing a survivor — a player may hold a correct reading indefinitely while everyone aboard still knows nothing. This is a distinction, not an extra gate: where a scene's own design is to show the evidence, the showing inside that scene is the informing, and no separate prior step is required. A control responding is not proof of identity or history — A-REPLY shows that something answers, never who. Physical network completion is not an ending theory — objective 12 is a walkable fact independent of objective 10, rescue, treatment and any ending; what the core contains is S4, and neither settles the other. Transaction, storage and display rules are not restated here.

Optional help is real, and no chain depends on it. L1–L4 are solvable with no person aboard. Yeva may name the crew roads and Dorn may describe crawl-mark use, each only when shown a record, each attributed with its basis, and neither reads an alien script or knows current topology. Help is not withheld to protect an intended deduction.

4. The two dependency orders

These are Regional25's two viable orders. The arrows are dependency priorities, not corridors; current access can interleave eligible objectives, and either order can defer R-3 until its short geometry exists.

Build-and-ship first — 1 → 2 → 3 → 4 → 8 → 6 → 10, then 5 → 7 → 9 → 11 → 12. Objective 1 supplies A-INSTALLED at R02/R03 and A-BERTH; the first death makes routes historical while those facts stay current, which is S1's whole opening. Objectives 2, 3, 4 and 8 earn movement, impact and positional parts and expose the physical wall/panel/slab differences used by L2 at 1 and 8, before systematic language comparison. Seeing those differences does not automatically grant their interpretation. Objective 6 completes without language, from gauges and the long access, so A-R18SEAL, A-CYCLE and D1 are available before L2's shortcut is understood; 10 then exposes the conditional apparent-departure preparation. Only afterwards do 5, 7 and 9 supply A-RUNS, A-TAAN, A-TRANSFER, A-AB, A-ADDRESS and A-REPLY, which reinterpret evidence already stored rather than unlocking it. Objective 11 asks for the same physical pressure-equalisation action it always asks for; the language routes make it predictable, never optional. Objective 12 depends on authored placement and a selected joining rule — not on 10, rescue, treatment or an ending.

Language-and-network first — 1 → 5 → 7 → 3 → 6 → 9 → 4 → 8 → 11 → 2 → 10 → 12. Riskier observation comes early: 5 and 7 give L1/L3 before most parts exist, and tested predictions shorten 6 and 9. Objective 4 arrives late, so R-3 and its rescue wait for a current short geometry — and no chain stalls, because 5, 7, 9 and 11 never require a survivor. Objective 11's iris is reached before the ship assemblies are complete; the physical action is identical. A-STRIP is therefore available early here and late in the other order, which changes when the three-record chronology can be assembled but not what it supports.

What differs, and what does not. No chain changes hands and no evidence changes meaning between the orders; what changes is when a record becomes available to compare and how much reinterpretation happens after the fact. Rescues and survivor help are optional in both.

Yeva-first versus Aster-first is a compact subcase, not a substitute for either order. It selects which person occupies R13 and which takes the R19 annex or R22 booth, and therefore which follow-up scene appears early. It sits inside either dependency order and changes no objective sequence, no evidence and no chain.

5. Four contrary cases

1 — Retained evidence, changed access. A-WALLMARK from R17 survives death with its place and life; the route to R17 is historical. The mark's meaning is unaffected, but whether that wall is passable now is current-only and needs a fresh visit.

2 — Wrong interpretation, concrete correction. From two A-TRANSFER pairs the player reads the recurring token as source. The corrective third pair at R23 shows the same token at a receiving end. The correction is physical and self-administered: nothing says "wrong", the earlier note keeps its original wording in revision history, and which continuation to investigate changes on the player's own reading.

3 — Absent versus uninformed participant. A-PROFILE is committed. If Dorn was never rescued, Y3 has no Dorn variant and nothing is lost invisibly. If Dorn is aboard but has not been shown the comparison, he is present and uninformed — a different state, and his knowledge does not update by proximity.

4 — Apparent versus deeper ending, with the departure branch unselected. Objective 10's assemblies can make readiness true with no language, survivor or core answer. The deeper transaction at R40 is separate and does not require it. The departure branch is an unselected owner choice, not a settled fact: Story30 recommends an irreversible apparent-credit departure — leaving the field ending revival and closing station exploration in this continuity — against the alternative of keeping readiness as a repair milestone with no departure at all, and the outline treats field-limited revival and irreversible departure as optional assumptions. Save policy is a further open question, not the whole of it. This article selects neither, and objective 12 neither triggers nor blocks either.

6. Integration handoff

Quantities, no longer wholly open. Mechanical35 is accepted and supplies the determinate endpoints: 10 C shared carrying capacity, a 12 O integral suit tank outside C/A, an ordinary-full canister of 8 O in a 2 C vessel and a prepared-full of 12 O in the same vessel, with conserved incremental transfer. What remains authored per route is the full-attempt return mark itself — a per-route value, not a missing model.

Nothing here requires a parent edit to be correct, because this article changes no rule; it records what the existing pages say and where they disagree. Two optional tags are the only parent changes named, and no tagging cascade is proposed to spread these aliases.

Coverage and limits

Complete for what was assigned: all 8 chains and all 15 primary scenes mapped, both Regional25 dependency orders traced through their actual evidence and state gates, and the evidence ledger deduplicated across them. Not covered, and not claimed: full occurrence/branch/failure variants per chain, exhaustive permutations of order and participation, any glyph or string catalogue, and the internals of anything owned by Mechanical35 beyond the endpoints cited. Ten encounters, two setpieces and thirteen mechanisms are referenced only through their owning rooms; this article adds none.

Nothing here establishes that the chains are discoverable, that the corrections land, or that any scene is affecting. Those need later owner-authorised implementation and observation, and none has occurred.