Natural temple and arena repairs — October 20, 2026

The whole v96p mod compiles and packages, but its strengthened natural arena test finds a real generation-order defect: a Runandor tree writes a runic log into the arena after room placement. The failed native receipt remains failed. The earlier v96o3 run independently found stone occupying the sparse template's intended room voids; the different successful v96m2 seed did not establish universal room clearance.

The current v96r2 candidate retains all 2,149 original ordered, typed block entries, the original altar and both guardian cages, the supported station, and 1,944 explicit air entries in the two original room interiors. Only this canonical arena's native factory and matching structure resource move after vegetation. A second actual failure after that stage change still found a runic log in a room, so stage order alone is insufficient across native chunk generation. The new Runic tree adapter protects native arena-start bounds during world generation, including the large tree’s separate direct soil write. Every original variant and random call remains; writes outside those bounds and manual tree growth retain their original behavior. The small per-tree native-start cache and thread context are released in finally. Every other structure keeps its generation order. Existing chunks are unchanged. All eight changed complete Java owners pass diagnostics against all 123 target dependencies. The complete normal native server now passes. Five independent verifiers bind the whole arena, four restored worlds, twenty-six portals, all five natural temples and the 800,000-chunk placement census. The actual native arena fixture verifies all 49,128 rotation/mirror positions, all 4,094 final natural positions, 8,188 blocked feature writes, 1,569 adjacent-chunk positions, and all five complete original tree generators. Outside writes, manual growth, nested context cleanup and prior block/entity-tag restoration pass. Current-cohort client and normal production acceptance remain pending.

Before any expanded-arena cohort was accepted, review also caught a retained literal loop bound in the controlled fixture. It now captures and restores every declared template position and reports its actual captured count; the earlier failed cohorts are not accepted as full expanded-arena proof.

The natural probe now verifies the actual native generation step and compares every one of the 4,094 final arena positions after FULL chunks, plus all twelve complete mirror/rotation placements. It does not skip trees or accept partial geometry. Its native feature fixture additionally checks both overwrite choices at every arena position, adjacent-chunk coordinates, allowed outside writes, manual choices, context cleanup, and all five complete original tree generators with restored block/entity-tag snapshots. The complete natural arena is checked again after neighbouring chunks and these controls. The connected-player client continues to require all sixty original temple cages, actual admitted mobs, original player state restoration, and explicit underwater movement/breathing/save-choice checks. The earlier twenty-six cage failures are not relabeled as fixed before the complete client passes.

Enderloom's nested source-admission validators now share one complete-read verification scope. Every distinct physical authority is fully hashed within each admission, changed metadata requires a fresh complete read, and final stability checks cover changes after the last consumer access. No cache survives admission; missing physical/archive-backed records keep their original verifier. Eight new generated controls and six existing lease controls pass. The real SDK refresh avoids 13,221,391,396 duplicate bytes; the full candidate admission avoids 41,861,131,203 duplicate bytes. These measure validation I/O, not Minecraft FPS.

Original, reusable verification implementations are published under tooling/native-verification. They accept the caller's complete byte verifier and preserve the full original record. The original native protection adapter and insertion-only recipe are also retained for review. The recipe requires the exact licensed original source; no original AoA source/assets or modified JAR is included.

Exact scope and immutable receipt hashes. Current playable profiles remain v93. Current candidate client/production, complete survival, optional integrations, multiplayer, migration, representative performance and the full historical/latest union remain open. The full-merge warning remains valid.

Current connected-player and native feature status

The accepted v96r2 server is not current client acceptance. Its actual connected-player child cfd030dc-7055-4bcc-b161-543291746614 / build 6193c845-77af-4c78-b7ee-89327a3a81b2 remains failed: 56 of 60 cages activated, admitting 112 mobs (Abyss 12/12, Barathos 12/12, Greckon 12/12, Haven 8/12, L’Borean 12/12). Actual native master volume was zero before content. All 32 admitted Borean mobs passed the controlled breathing/navigation/controller and absent/false/foreign-string save checks. This does not establish combat or survival parity. Read-only native saved-region inspection finds oak logs and coloured Haven leaves across all four central Sceptron spawn volumes; original 2.2-by-2.2 mob dimensions and native collision checks remain unchanged.

Candidate v96s adds region-lifetime concurrent bounds caching on the real WorldGenRegion and narrow native Minecraft tree occupancy/write hooks for the arena and five temple identities. Complete original native tree configurations and algorithms remain; manual ServerLevel growth and outside writes retain their normal path. All source diagnostics compile. The complete native server 9a16623a-f6c2-4d4a-b91a-92cdac88fe88 / build 3380efd1-3d78-4d58-83fb-68ada8bc1a97 then fails a different complete natural Greckon comparison: lab_brick expected at (22319,55,-15996), cave_air observed. It is not accepted or promoted.

The current v96t1 source candidate extends the same actual native bounds cache with feature-volume intersection. The complete existing BiomeFriendlyLake source gains an insertion-only exclusion for overlapping natural lake candidates before carving/fluid writes, alongside its existing village exclusion. It retains every original algorithm/configuration elsewhere and explicit manual placement. Six changed complete sources compile against all 123 native dependencies into eight classes. Its complete normal server run is pending, with actual native protected/outside/manual lake fixtures and all inherited worldgen/temple/arena checks. One probe flag omission was rejected before intake/build; the corrected probe retains the complete inherited world-generation workload. No gameplay acceptance is inferred from diagnostic compilation.

Storage during this continuation: 30 closed B: compiler/runtime copies (994,935,972 logical bytes) and 15 closed E: copies (497,522,702 bytes) were retired after current source/SDK/proof pinning and idle-state checks. All 5,614 protected physical files and 65 archive-backed paths verify afterward. No new recovery archive or unrelated-drive files were created or changed. Future Studio outputs use only the already owned E: development directory; existing player profiles, source authorities and playable products remain. The separate aggregate C: 20GB budget is still unmet. Current playable profiles remain v93; the full-merge warning remains valid.

Accepted complete native feature-coexistence server

The corrected v96t3 complete native server is accepted (receipt 399dc22f-4be3-4bcd-ac70-028e0355db02, build 38ff4f86-81ce-4fe1-a6cc-d85370871da6). Product: 151,056,273 bytes / 25,170 members, SHA-256 16818b3d0d67c72cf0b706962e445759e170734d2906e6f6e67a3998cd0c9ad3. All five natural temples retain all 23,410 authored blocks and sixty native cages. The full 4,094-position natural arena and all 49,128 transformed positions pass. Real native Minecraft tree predicate/writer protection passes at every arena position, with outside/manual controls. The complete original lake rejects an overlapping 2,048-cell candidate without mutation, while outside and manual native calls each perform 251 carving/fluid writes; their original terrain and tile tags are restored. Six independent server verifiers pass, including 524,288 restored-world positions, 26 portals/25 entity round trips, the complete temple/guardian/projectile cohort, and the 800,000-chunk spacing census. Twenty-four generated arena faults are rejected. This is scoped server evidence; all-sixty-cage client, current-cohort visuals, normal production/Hari, survival, multiplayer and full historical/latest union remain open. Current player profiles remain v93.

The earlier v96t1 preflight rejected three draft additions incorrectly expressed as extensions of an accepted baseline that did not yet contain them. v96t3 admits their complete current bytes as additions, retains their exact unaccepted draft history and provenance separately, and keeps all accepted-source extension checks. The whole 1,225-recipe merge passes (1,218 retained, three extended, four added). No world or mod build was created by the rejected preflight. Original owned region-cache/tree adapters and the lake insertion-only recipe are now available under development/adapters/forge-1.20.1/structure-feature-protection; original AoA source/assets and modified JARs are not republished.

Accepted all-sixty-cage client and production mapping follow-up

The complete v96t3 client is accepted: native receipt 0962091f-1766-4358-a1b1-52b095c2c318, build 5305c2c8-b3dc-4c89-9aa8-fb4e87954ec1. All 60 naturally generated cages activate with the actual connected disposable player, admitting 132 mobs: Abyss 23, Barathos 32, Greckon 27, Haven 21 and L'Borean 29. All 87 underwater save-choice cases pass. The player, original settings and fixture state are restored. Actual native master volume is zero before content; the original six cage textures, eight portal GPU draws and four restored-world effects pass. Independent verification rejects twelve corrupted cage captures, and all 25,170 packaged server/client members compare equal except the main build timestamp. This is controlled native activation/visual evidence, not a full survival playthrough.

The same cohort's ordinary production server correctly fails before world creation: the packaged Mixin configuration selects the old 13-owner refmap, omitting SnapshotRealmTimeMixin's native timeOfDay mapping. The next v96u candidate changes only that config selection to the already generated 16-owner refmap. Every old mapping and namespace is retained exactly; every Java recipe hash is unchanged. The old refmap also remains packaged. Three generated mapping corruptions are rejected.

The v96u whole compile, annotation processing and packaging pass, but the complete native scenario fails portal entity admission. A diagnostic replay using the actual captured VM recipe, unchanged compiled inputs and the exact original failed world seed passes the full server workload. The earlier whole-build receipt remains failed. Temporary portal-config choices were writing through Forge's autosaving file provider, triggering asynchronous reloads. A repaired separate probe uses a complete typed/comment-preserving scoped in-memory provider through Forge's public API; its exact-failed-seed diagnostic passes all 26 owners and 25 entity round trips, with every original choice, comment, provider identity and disk byte unchanged. It retains all disabled-travel/cooldown assertions and unchanged production portal code. A fresh normal whole build is running; diagnostic replay is not whole-build admission. Normal production/Hari, survival, multiplayer and full historical/latest union remain open. Player profiles remain v93.

Accepted production mapping cohort (v96u3)

The corrected whole server (fc075431-d1e1-4ea9-8d4a-9d56992ec1b9 / build 3177253d-8609-4f44-ac1b-0c7ba3ed4276) and muted client (3ca42061-31fb-4143-9d8e-a30ad536c1b5 / build c045444a-1d82-4712-9b24-ffdcc79ffaf6) pass. All sixty natural cages activate, admitting 121 mobs in this native seed; all 81 underwater saved-choice cases pass. The earlier cohort's different randomized mob count is retained as its own observation. All 25,170 server/client members compare equal except the main build timestamp. The packaged change from T3 retains every member byte except the exact refmap selector and that timestamp, with four generated corruptions rejected. Canonical product: 151,056,284 bytes, SHA-256 dcbe80e9f82a625e2b637ce6e67887906513d2f78058d0099a1c20b30ce33370.

The same exact canonical product then passes ordinary production startup, forty native ticks, save and restart. Its signed-in required-Hari62 client passes all twenty-four actual native GUI draw/state cases and movement, with master audio zero before content. The independent current renderer verifier binds both native backends, the complete products, 75 unchanged production-world files, original choices and disposal of the temporary client. It does not certify representative performance, the 2,000 FPS goal or full survival. The existing owned renderer QA profile is updated; ordinary player profiles remain v93. No additional runtime/library installation was copied.

A read-only native logging observer now captures all eight real error stacks across the whole asynchronous workload without replacing game/logging methods or removing existing appenders. The earlier root-only observer missed events and remains diagnostic-only, not zero-error proof. The corrected trace identifies empty Brain compounds in ten Opteryx nest payloads and stale authoring-world attachment coordinates in ten decorations across eight template owners. The next candidate uses Forge's entity-processing placement API to retain every original template/asset byte and entity while preparing correct target-native placement data. Existing chunks/entities are outside that processor. Native acceptance of this next fix remains pending; save migration, multiplayer, optional integrations, sustained performance and the full historical/latest union remain open.

Complete template entity repair candidate

All three changed complete Java owners pass diagnostics in the 1,227-recipe candidate. The actual corpus contains thirty-four entities: twenty affected payloads and fourteen unaffected residents. The initial native fixture incorrectly treated every resident as affected; that failed run remains failed (cc0252fd-21fa-431a-abc9-5bc7c38948ca / f7f8c67c-266a-4ced-9ffa-34d391cb5207). The corrected fixture covers all 816 mirror/rotation/pivot cases and compares every typed NBT field, preserving resident data as well as fixing placement. The same original -3329602126742680344 world seed passes a full diagnostic replay with all 816 cases and zero observed ERROR/FATAL events across the asynchronous world/entity/item workload; the unchanged observer previously captured eight actual errors. Runtime classpath/mappings remain exact before/after. This diagnostic replay is not fresh whole-build acceptance or a full survival/multiplayer claim. The new normal whole server/client and production cohort remain pending.

Accepted complete template entity repair cohort (v96v2)

The separate whole server (f088d240-94e9-498e-a911-ba00f0d7e421 / 8498ee3f-31f2-465f-af8f-cdddd05cb389) and muted client (fb2c71b8-23d2-4a6c-8741-ee20d0b386ef / 1b114eee-faa7-4ac8-912d-5f916b701774) pass. The 1,227-recipe product has 25,171 members, 151,058,907 bytes, SHA-256 f4a741a2b1468c27cb1ca9d5831d10ab9d38772485fbea5ba88c808fa9ac54e1. All 816 complete entity cases pass with every other typed NBT field retained; all original template bytes remain exact. The unchanged full-scope observer captures zero errors in both the fresh native workload and the exact original failing-seed replay. Independent comparison binds the replayed program bytes to the fresh accepted product, apart from build time. Every other packaged member is exact outside the reviewed owner/pool changes and eight shifted nested-builder debug line tables; all executable builder bytes remain exact.

All sixty natural cages activate in the client, admitting 130 mobs in this seed. All 78 underwater saved-choice cases pass. The entire server/client archive compares equal except build time. All inherited arena, temple, spacing, restored-world, guardian and portal checks pass again. Actual native master volume is zero.

The same canonical product passes fresh ordinary startup, debug ticks, save and restart in the existing owned production runtime. All dependency libraries and the previous production world remain exact; no full runtime/library copy was made. A read-only path check initially rejected a volume-relative root before any game or product/world mutation; the normalized native-path retry passes and retains the failure. The required-Hari62 client then passes all 24 native GUI cases, movement, actual master zero, source-choice preservation and disposable-client cleanup. Independent renderer proof binds all 75 fresh source/imported world files and the old worlds.

The original processor and integration instructions are published under development/adapters/forge-1.20.1/template-entity-data; original AoA source/templates/assets and a modified JAR are not republished. The renderer QA profile is updated; ordinary player profiles remain v93 pending further qualification. The first real dedicated-client attempt admitted the player but its harness waited on the finalized telemetry path instead of the live sandbox stream. Its failure and cleanup remain recorded; the corrected retry reuses the same owned test world. Multiplayer, old-save acceptance, optional full-pack integration, sustained performance and the complete historical/latest union remain separate open gates.

Actual dedicated client, prior save and playable promotion

The same v96v2 product now admits the signed-in player to a real dedicated server in all 26 AoA dimensions. Each observation requires a cleared loading screen and fresh native frames. Twenty-four native GUI cases pass under required Hari62 Vulkan, at actual master volume zero. The complete observed inventory is identical after dedicated-server restart and rejoin. The 26 screenshots establish admission and sky/fog capture, not complete terrain raster parity or natural portal acquisition. Six generated broken controls are rejected.

A disposable copy of the actual preceding Current Playtest world is then read by the new product: all 2,025 allocated source chunks reach the native loaded condition, followed by normal ticks, save and restart. The original 69 source files and prior mod remain unchanged. The copied save opens in the muted required-Vulkan client and passes 24 GUI cases. An earlier attempt really loaded the chunks but failed its harness because the rolling console lost markers; that failure remains, and acceptance uses the complete owned server log with unique markers on two fresh runs. This proves this particular save, not arbitrary Minecraft-version migration.

Both existing Current Playtest profiles are now updated in place from v93 to v96v2: 151,058,907 bytes, SHA-256 f4a741a2b1468c27cb1ca9d5831d10ab9d38772485fbea5ba88c808fa9ac54e1. All earlier registry keys and all 24,295 prior archive paths remain in the 25,171-member product. The 69 ordinary and 732 Hari-profile user files remain byte-identical; no profile, world backup or runtime/library copy was added. User-edited notes and launch choices remain. Vanilla hearts remain default.

The whole historical/latest union, Ancient Cavern/Immortallis reconciliation, remaining original consumers, natural survival progression, arbitrary save migrations, full-pack optional compatibility, complete visual parity and sustained 2,000 FPS remain open. The complete-merge warning is not cleared. Current evidence supersedes the earlier “ordinary profiles stay v93” and “all network/save acceptance remains open” statements only within the exact scopes above.