feat(log): log authority is the fleet default — adopt-at-open, oracle-gated; plus the power-cut throw-site cures and the loud torn-record contract

THE DEFAULT FLIP (ruled on proven evidence — at-ack survived 301/301
acked-writes-through-power-cut in block-layer fault injection; deferred
tree authority demonstrably loses flush-covered acks): a brain with NO
stored authority artifact now ADOPTS LOG AUTHORITY AT OPEN. The oracle
gates the flip exactly as the guarded adoption path always did — curable
divergences baseline-backfilled, the flip lands ONLY on a green verdict —
and a brain that cannot verify STAYS tree-authoritative loudly, with the
refusal recorded on the switch artifact so subsequent opens are cheap.
config logAuthority: 'defer' is the explicit documented opt-out (no
automatic adoption; declared flush-window loss; adoptLogAuthority() flips
later). A stored artifact always wins. RELEASES.md carries the posture.

Two standing .fails debt pins FLIP TO HOLDING under the default: the
at-ack crash-survival gap and the ack-at-log durability target — both now
permanent asserted truths, not aspirations.

POWER-CUT THROW SITES (fault-injection findings, brainy-alone config):
- A manifest-listed-but-unloadable column segment QUARANTINES at
  discovery (loud once, counted always, quarantinedSegments() exposed for
  the heal) and the field serves its remaining segments DEGRADED — never
  a raw throw killing every query on the field. Real storage faults still
  propagate untouched.
- Torn generation artifacts (NaN/garbage in manifest or counter) DISCARD
  with narration at the store's open and recovery re-derives — plus a
  defensive finite-integer guard at the init consumer. Never a RangeError
  killing an open.

THE LOUD TORN-RECORD CONTRACT: an existing-but-unparseable stored record
now surfaces as a typed, counted TornRecordError on every entity-read
surface (including fifteen previously-blind per-item batch catches);
ENOENT stays clean-absent; artifact readers with designed absent-recovery
keep null-tolerance behind the loud floor. Disk corruption can no longer
read as silent data invisibility.

Suite migration: the default's pins inverted deliberately, generation
baselines made relative, quarantine-contract pins rewritten to the ruled
behavior.

Gates: tsc 0 · unit 2065/2065 (159 files) · integration 826 (93 files) ·
conformance 31/31 · kill-matrix 15/15 · torn-open guards 2/2.

RELEASES.md

@@ -37,6 +37,17 @@ The theme: **writes ack fast and honestly, startup adopts instead of rebuilding,
 every query path serves, announces, or refuses — never silently degrades.** Ships as
 one release together with the matching native accelerator version.
 
+**The storage-authority posture (the release's headline):** a NEW brain's default
+is **durable-at-ack log authority** — the generation log is the source of truth,
+every write acknowledgment is covered by a group-committed fsync, and crash
+recovery is a replay of the log (an acked write survives power loss, proven by
+fault-injection tests). An EXISTING brain adopts at its first open under 10.0.0,
+gated by a verification oracle: the log is replayed and diffed against stored
+truth record-by-record; curable gaps are backfilled; the brain flips only on a
+green verdict and a brain that cannot verify stays on the previous posture and
+says so loudly. The explicit opt-out is `logAuthority: 'defer'` in the config
+(no automatic adoption; flip later with `adoptLogAuthority()`).
+
 **Why a major:** the generation log gains write format v2 — new segments carry typed,
 versioned records with integrity seals. A 9.x build refuses a v2 segment with a clear
 version-naming error (never a misread), which means **a brain written by 10.x cannot

src/brainy.ts

@@ -202,6 +202,7 @@ import {
   flipToLogAuthority,
   recordDigest,
   nounEntityTruth,
+  LOG_AUTHORITY_PATH,
   type LogAuthorityRecord,
   type LogAuthorityStorage,
   type OracleReport
@@ -1371,7 +1372,20 @@ export class Brainy<T = any> implements BrainyInterface<T> {
       // gap for observability.
       for (const provider of this.versionedIndexProviders()) {
         const providerGen = provider.generation()
-        const committed = BigInt(this.generationStore.committedGeneration())
+        // Defensive finite-integer guard: committedGeneration() is validated
+        // at the store's open (torn artifacts discard, narrated) — but a
+        // RangeError here would kill the whole open, so the consumer guards
+        // too. A non-finite value narrates and skips the gap check (the
+        // provider's own replay contract still governs).
+        const committedRaw = this.generationStore.committedGeneration()
+        if (!Number.isSafeInteger(committedRaw) || committedRaw < 0) {
+          prodLog.warn(
+            `[Brainy] committed generation is non-integer (${String(committedRaw)}) at ` +
+              `init — torn-artifact survivor; skipping the provider replay-gap check`
+          )
+          continue
+        }
+        const committed = BigInt(committedRaw)
         if (providerGen < committed) {
           prodLog.info(
             `[Brainy] Versioned index provider is at generation ${providerGen} ` +
@@ -1492,16 +1506,58 @@ export class Brainy<T = any> implements BrainyInterface<T> {
         this._generationStampingActive = true
       }
 
-      // LOG-AUTHORITY SWITCH (checked at open only): a brain that has
-      // flipped to log-authoritative storage gets durable-at-ack fact
-      // writes (group-committed fsync covering every ack). Default 'tree'
-      // = today's behavior, zero added latency.
+      // LOG-AUTHORITY SWITCH (checked at open only). A STORED artifact
+      // always wins: an already-flipped brain runs durable-at-ack; an
+      // explicitly-recorded tree posture is honored. With NO artifact, the
+      // 10.0.0 FLEET DEFAULT is ADOPT-AT-OPEN (config logAuthority:
+      // 'adopt'): the verification oracle gates the flip — curable
+      // divergences are baseline-backfilled, the brain flips ONLY on green,
+      // and a brain that cannot go green STAYS tree-authoritative LOUDLY
+      // with the refusal recorded (cheap subsequent opens; an operator
+      // re-runs adoptLogAuthority() after fixing the divergence).
+      // 'defer' is the documented opt-out: no automatic adoption.
       if (!this.isReadOnly) {
+        const storedArtifact = await this.storage
+          .readRawObject(LOG_AUTHORITY_PATH)
+          .catch(() => null)
         const authority = await readLogAuthority(this.storage)
         this._logAuthority = authority
         if (authority.authority === 'log') {
           this.generationStore.setLogDurability('at-ack')
           prodLog.info('[Brainy] storage authority: generation log (durable-at-ack enabled)')
+        } else if (
+          storedArtifact === null &&
+          this.config.logAuthority === 'adopt' &&
+          this.generationStore.getFactLog() !== null
+        ) {
+          try {
+            await this.adoptLogAuthority()
+            prodLog.info(
+              '[Brainy] storage authority adopted at open: generation log ' +
+                '(fleet default; oracle green; durable-at-ack enabled)'
+            )
+          } catch (err) {
+            // The guarded ruling: a brain that cannot verify STAYS tree,
+            // loudly, with the refusal recorded so subsequent opens are
+            // cheap. Never a silent half-state; never a failed open.
+            const reason = (err as Error).message
+            prodLog.warn(
+              `[Brainy] log-authority adoption REFUSED at open — this brain stays ` +
+                `tree-authoritative until an operator resolves the divergence and ` +
+                `re-runs adoptLogAuthority(). Reason: ${reason}`
+            )
+            try {
+              const refusal: LogAuthorityRecord = {
+                authority: 'tree',
+                adoptRefusal: { at: Date.now(), reason: reason.slice(0, 500) }
+              }
+              await this.storage.writeRawObject(LOG_AUTHORITY_PATH, refusal)
+              this._logAuthority = refusal
+            } catch {
+              // Unrecordable refusal = the next open retries the oracle —
+              // the conservative outcome.
+            }
+          }
         }
       }
 
@@ -15786,7 +15842,8 @@ export class Brainy<T = any> implements BrainyInterface<T> {
       force: config?.force ?? false,
       // Engine-owned persistence cadence — defaults resolve at the trigger
       // site (policy 'auto': 512 writes / 30s interval / 2s idle).
-      persistence: config?.persistence
+      persistence: config?.persistence,
+      logAuthority: config?.logAuthority ?? 'adopt'
     }
   }
 

src/db/generationStore.ts

@@ -468,9 +468,26 @@ export class GenerationStore {
       | null
     const manifest = (await this.storage.readRawObject(MANIFEST_PATH)) as GenerationManifest | null
 
-    this.committed = manifest?.generation ?? 0
-    this.horizonGen = manifest?.horizon ?? 0
-    this.counter = Math.max(counterFile?.generation ?? 0, this.committed)
+    // TORN-ARTIFACT VALIDATION (power-loss survivors): a torn manifest or
+    // counter can carry NaN/garbage where a generation belongs — unguarded,
+    // that NaN reaches BigInt() conversions at init and kills the open with
+    // a RangeError. A non-finite-integer generation is DISCARDED with
+    // narration (the conservative floor: 0 = re-derive from the record
+    // directories / fact log below, exactly the recovery machinery's job).
+    const finiteGen = (v: unknown, source: string): number => {
+      if (typeof v === 'number' && Number.isSafeInteger(v) && v >= 0) return v
+      if (v !== undefined && v !== null) {
+        prodLog.warn(
+          `[GenerationStore] ${source} carries a non-integer generation ` +
+            `(${String(v)}) — torn write survivor; discarding and re-deriving ` +
+            `from recovery (never a RangeError at open)`
+        )
+      }
+      return 0
+    }
+    this.committed = finiteGen(manifest?.generation, 'manifest')
+    this.horizonGen = finiteGen(manifest?.horizon, 'manifest horizon')
+    this.counter = Math.max(finiteGen(counterFile?.generation, 'generation counter'), this.committed)
 
     // Discover existing generation record directories.
     const recordPaths = await this.storage.listRawObjects(GENERATIONS_PREFIX)

src/db/logAuthority.ts

@@ -43,6 +43,12 @@ export interface LogAuthorityRecord {
     nounsChecked: number
     verbsChecked: number
   }
+  /**
+   * Recorded when an OPEN-TIME adoption attempt (the 10.0.0 fleet default)
+   * was refused — the oracle could not go green. Keeps subsequent opens
+   * cheap; an operator re-runs adoptLogAuthority() after resolving it.
+   */
+  adoptRefusal?: { at: number; reason: string }
 }
 
 /** The narrow storage surface this module needs. */

src/index.ts

@@ -362,6 +362,15 @@ export { MemoryStorage, createStorage }
 // FileSystemStorage is exported separately to avoid browser build issues.
 export { FileSystemStorage } from './storage/adapters/fileSystemStorage.js'
 
+// Torn-record surface: a stored file that EXISTS but cannot be decoded throws
+// a typed, catchable error on entity reads (never a silent "not found"), and
+// every encounter is counted on a per-process gauge.
+export {
+  TornRecordError,
+  isTornRecordError,
+  getTornRecordGauge
+} from './storage/tornRecordError.js'
+
 // Export types
 import type {
   Vector,

src/indexes/columnStore/ColumnStore.ts

@@ -31,6 +31,7 @@ import { ColumnSegmentCursor, TailBufferCursor, type CursorEntry } from './Colum
 import { writeSegmentToBuffer, readSegmentFromBuffer } from './ColumnSegmentFormat.js'
 import { RoaringBitmap32 } from '../../utils/roaring/index.js'
 import { compareCodePoints } from '../../utils/collation.js'
+import { prodLog } from '../../utils/logger.js'
 
 /**
  * Configuration for the ColumnStore.
@@ -612,6 +613,24 @@ export class ColumnStore implements ColumnStoreProvider {
   /**
    * Get all segment cursors for a field, loading from storage if needed.
    */
+  /**
+   * Per-field quarantine ledger for torn segments (power-loss survivors:
+   * manifest-listed but unloadable). A quarantined segment is skipped with
+   * per-doubling narration and the field serves its REMAINING segments as a
+   * DEGRADED-ANNOUNCED result — never a raw throw killing the query, never
+   * a silent drop. Cleared when a heal/rebuild rewrites the field.
+   */
+  private readonly segmentQuarantine = new Map<string, { error: string; hits: number }>()
+
+  /** Torn-segment quarantine entries for a field (observability + heal input). */
+  quarantinedSegments(field: string): Array<{ segment: string; error: string; hits: number }> {
+    const out: Array<{ segment: string; error: string; hits: number }> = []
+    for (const [key, q] of this.segmentQuarantine) {
+      if (key.startsWith(`${field}:`)) out.push({ segment: key.slice(field.length + 1), error: q.error, hits: q.hits })
+    }
+    return out
+  }
+
   private async getSegmentCursors(field: string): Promise<ColumnSegmentCursor[]> {
     const manifest = this.manifests.get(field)
     if (!manifest) return []
@@ -622,11 +641,38 @@ export class ColumnStore implements ColumnStoreProvider {
       let cursor = this.segmentCache.get(cacheKey)
 
       if (!cursor) {
-        // loadSegmentCursor either returns a cursor or THROWS — a corrupt /
-        // missing manifest-listed segment raises ColumnSegmentLoadError and a
-        // real storage fault propagates, so a listed segment is never silently
-        // dropped from the result set.
+        const quarantined = this.segmentQuarantine.get(cacheKey)
+        if (quarantined) {
+          // Already-quarantined torn segment: skip, count, narrate per doubling.
+          quarantined.hits++
+          if ((quarantined.hits & (quarantined.hits - 1)) === 0) {
+            prodLog.warn(
+              `[ColumnStore] field '${field}' serving DEGRADED: torn segment ${seg.id} ` +
+                `quarantined (${quarantined.error}) — ${quarantined.hits} queries served ` +
+                `without it; heal/rebuild the metadata index to restore`
+            )
+          }
+          continue
+        }
+        try {
           cursor = await this.loadSegmentCursor(field, seg)
+        } catch (err) {
+          if (err instanceof ColumnSegmentLoadError) {
+            // POWER-LOSS SURVIVOR: a manifest-listed segment whose bytes are
+            // torn/absent. Quarantine at DISCOVERY and serve the remaining
+            // segments degraded-announced — a raw throw here killed every
+            // query on the field forever; a silent skip hid the loss. The
+            // quarantine is the middle: loud once, counted always, healable.
+            this.segmentQuarantine.set(cacheKey, { error: (err as Error).message, hits: 1 })
+            prodLog.error(
+              `[ColumnStore] torn segment QUARANTINED at discovery: field '${field}' ` +
+                `segment ${seg.id} — ${(err as Error).message}. The field serves its ` +
+                `remaining segments DEGRADED until a heal/rebuild rewrites it.`
+            )
+            continue
+          }
+          throw err // real storage faults propagate — never absorbed
+        }
         this.segmentCache.set(cacheKey, cursor)
       }
 

src/storage/adapters/fileSystemStorage.ts

@@ -18,6 +18,11 @@ import {
 } from '../baseStorage.js'
 import { getBrainyVersion } from '../../utils/index.js'
 import { isAbsentError } from '../../utils/errorClassification.js'
+import {
+  TornRecordError,
+  isUnparseablePayloadError,
+  registerTornRecordEncounter
+} from '../tornRecordError.js'
 
 // Node.js modules - dynamically imported to avoid issues in browser environments
 let fs: any
@@ -410,8 +415,22 @@ export class FileSystemStorage extends BaseStorage {
   /**
    * Primitive operation: Read object from path
    * All metadata operations use this internally via base class routing
-   * Enhanced error handling for corrupted metadata files (Bug #3 mitigation)
    * Supports reading both compressed (.gz) and uncompressed files for backward compatibility
+   *
+   * Read contract (loud errors, never quiet losses):
+   * - Genuine absence (ENOENT on every variant) → `null`. Only a missing file
+   *   is "not found".
+   * - TORN record (a file EXISTS but its bytes cannot be decoded — invalid
+   *   JSON, truncated/garbled gzip) → the encounter is registered (production
+   *   ERROR log + per-process gauge) and a typed {@link TornRecordError} is
+   *   thrown. Corruption must NEVER read as absence: callers that can degrade
+   *   (manifest recovery, rebuildable statistics) catch the typed error at
+   *   their sites; entity reads surface it.
+   *   Legacy dual-format exception: when the `.gz` variant is torn but the
+   *   uncompressed fallback decodes, the recovered object is returned — AFTER
+   *   the torn `.gz` was logged and counted (loud recovery, not a silent skip).
+   * - Real storage fault (EIO/EACCES/EMFILE/…) → propagates as itself; a
+   *   fault is neither absence nor corruption and must not be reshaped.
    */
   protected async readObjectFromPath(pathStr: string): Promise<any | null> {
     await this.ensureInitialized()
@@ -419,7 +438,10 @@ export class FileSystemStorage extends BaseStorage {
     const fullPath = path.join(this.rootDir, pathStr)
     const compressedPath = `${fullPath}.gz`
 
-    // Try reading compressed file first (if compression is enabled or file exists)
+    // Try reading compressed file first (if compression is enabled or file exists).
+    // A torn .gz is remembered so the uncompressed fallback can either recover
+    // (legacy dual-format installs) or surface the corruption typed.
+    let tornCompressed: TornRecordError | null = null
     try {
       const compressedData = await fs.promises.readFile(compressedPath)
       const decompressed = await new Promise<Buffer>((resolve, reject) => {
@@ -430,9 +452,16 @@ export class FileSystemStorage extends BaseStorage {
       })
       return JSON.parse(decompressed.toString('utf-8'))
     } catch (error: any) {
-      // If compressed file doesn't exist, fall back to uncompressed
-      if (error.code !== 'ENOENT') {
-        console.warn(`Failed to read compressed file ${compressedPath}:`, error)
+      if (error.code === 'ENOENT') {
+        // No compressed variant — fall through to the uncompressed path.
+      } else if (isUnparseablePayloadError(error)) {
+        // The .gz EXISTS but cannot be decoded (zlib Z_* error or JSON
+        // SyntaxError after gunzip): torn record. Register NOW (log + gauge),
+        // then attempt the uncompressed fallback as a recovery read.
+        tornCompressed = registerTornRecordEncounter(`${pathStr}.gz`, error)
+      } else {
+        // Real storage fault on an existing .gz (EIO/EACCES/…): propagate.
+        throw error
       }
     }
 
@@ -442,24 +471,26 @@ export class FileSystemStorage extends BaseStorage {
       return JSON.parse(data)
     } catch (error: any) {
       if (error.code === 'ENOENT') {
+        // No uncompressed file. If the .gz variant existed but was torn, the
+        // object EXISTS and is unreadable — that must surface typed, never as
+        // "absent". Otherwise this is genuine absence.
+        if (tornCompressed !== null) {
+          throw tornCompressed
+        }
         return null
       }
 
-      // Enhanced error handling for corrupted JSON files (race condition from Bug #3)
-      if (error instanceof SyntaxError || error.name === 'SyntaxError') {
-        console.warn(
-          `⚠️  Corrupted metadata file detected: ${pathStr}\n` +
-          `   This may be caused by concurrent writes during import.\n` +
-          `   Gracefully skipping this entry. File may be repaired on next write.`
-        )
-        return null
+      // The file EXISTS but its content cannot be parsed: torn record.
+      // Register (production ERROR + gauge) and throw typed — a corrupt row
+      // must be distinguishable from a missing row, or nothing ever heals it.
+      if (isUnparseablePayloadError(error)) {
+        throw registerTornRecordEncounter(pathStr, error)
       }
 
       // A real storage fault (EIO/EACCES/EMFILE/…) is NOT "object absent". The
-      // ENOENT branch (above) already returns null, and the corrupted-JSON
-      // branch (above) is a deliberate concurrent-write tolerance; a genuine
-      // fault reaching here must propagate loudly rather than masquerade as a
-      // missing object — which would corrupt reads and drive needless rebuilds.
+      // ENOENT branch (above) already returns null; a genuine fault reaching
+      // here must propagate loudly rather than masquerade as a missing object
+      // — which would corrupt reads and drive needless rebuilds.
       throw error
     }
   }

src/storage/baseStorage.ts

@@ -32,6 +32,7 @@ import { BlobStorage, type BlobStoreAdapter } from './blobStorage.js'
 import { unwrapBinaryData } from './binaryDataCodec.js'
 import { prodLog } from '../utils/logger.js'
 import { isAbsentError } from '../utils/errorClassification.js'
+import { isTornRecordError } from './tornRecordError.js'
 import { BrainyError, ProtectedArtifactError, DerivedArtifactMissingError } from '../errors/brainyError.js'
 import { MetadataWriteBuffer } from '../utils/metadataWriteBuffer.js'
 import {
@@ -674,6 +675,10 @@ export abstract class BaseStorage extends BaseStorageAdapter {
           // — hash verification must run on the original content bytes.
           return unwrapBinaryData(data)
         } catch (error) {
+          // A TORN blob object (exists but undecodable) must not read as
+          // "blob absent" — that would misdiagnose disk corruption as a
+          // missing blob. Propagate the typed error to the blob layer.
+          if (isTornRecordError(error)) throw error
           return undefined
         }
       },
@@ -768,6 +773,20 @@ export abstract class BaseStorage extends BaseStorageAdapter {
       if (m) hashes.add(m[1])
     }
 
+    // Recovery-path read: a TORN object here maps to "not usable" (null) BY
+    // DESIGN — the adapter has already logged + counted the encounter, and
+    // treating a torn `_cas/` copy as absent lets the re-copy from `_cow/`
+    // OVERWRITE the corrupt file with the good original (the heal), while a
+    // torn `_cow/` original is reported via `incomplete`. Real faults propagate.
+    const readOrNullIfTorn = async (p: string): Promise<any | null> => {
+      try {
+        return await this.readObjectFromPath(p)
+      } catch (error) {
+        if (isTornRecordError(error)) return null
+        throw error
+      }
+    }
+
     let adopted = 0
     let alreadyPresent = 0
     let incomplete = 0
@@ -775,15 +794,15 @@ export abstract class BaseStorage extends BaseStorageAdapter {
       // A blob counts as present only when BOTH its bytes and its metadata
       // already live in `_cas/`. A half-adopted blob (bytes without meta — the
       // exact "Blob metadata not found" state) is re-adopted.
-      const casBlob = await this.readObjectFromPath(`_cas/blob:${hash}`)
-      const casMeta = await this.readObjectFromPath(`_cas/blob-meta:${hash}`)
+      const casBlob = await readOrNullIfTorn(`_cas/blob:${hash}`)
+      const casMeta = await readOrNullIfTorn(`_cas/blob-meta:${hash}`)
       if (casBlob !== null && casMeta !== null) {
         alreadyPresent++
         continue
       }
 
-      const cowBlob = await this.readObjectFromPath(`_cow/blob:${hash}`)
-      const cowMeta = await this.readObjectFromPath(`_cow/blob-meta:${hash}`)
+      const cowBlob = await readOrNullIfTorn(`_cow/blob:${hash}`)
+      const cowMeta = await readOrNullIfTorn(`_cow/blob-meta:${hash}`)
       if (cowBlob === null || cowMeta === null) {
         // Can't register a blob the store can't fully describe — report it so an
         // operator investigates rather than silently half-adopting.
@@ -1134,12 +1153,28 @@ export abstract class BaseStorage extends BaseStorageAdapter {
    * cache (record-layer files are written through
    * {@link BaseStorage.writeRawObject} only).
    *
+   * TORN-record contract (deliberate, loud-by-design): this surface serves
+   * SYSTEM ARTIFACTS — manifests with recovery paths, markers whose verdict
+   * machinery treats "unreadable" as rescan, generation/transaction records
+   * whose recovery is built for absent artifacts. For these readers a torn
+   * file maps to their existing absent-artifact degrade, so a typed
+   * torn-record error from the adapter is caught here and returned as `null`
+   * — AFTER the adapter has already logged a production ERROR and counted
+   * the per-process torn-record gauge (never silent). Entity reads do NOT go
+   * through this surface; they use the canonical read paths, which propagate
+   * the typed error. Real storage faults (EIO/EACCES/…) still propagate.
+   *
    * @param path - Storage-root-relative object path (e.g. `_system/manifest.json`).
-   * @returns The parsed object, or `null` if absent.
+   * @returns The parsed object, or `null` if absent (or torn — logged + counted).
    */
   public async readRawObject(path: string): Promise<any | null> {
     await this.ensureInitialized()
-    return this.readObjectFromPath(path)
+    try {
+      return await this.readObjectFromPath(path)
+    } catch (error) {
+      if (isTornRecordError(error)) return null
+      throw error
+    }
   }
 
   /**
@@ -2146,6 +2181,9 @@ export abstract class BaseStorage extends BaseStorageAdapter {
                 if (!metadata) return null
                 return { deserialized, metadata }
               } catch (error) {
+                // A TORN record must surface typed — a paginated read that
+                // silently skips a corrupt row hides data loss from the caller.
+                if (isTornRecordError(error)) throw error
                 // Skip nouns that fail to load
                 return null
               }
@@ -2175,6 +2213,8 @@ export abstract class BaseStorage extends BaseStorageAdapter {
           }
         }
       } catch (error) {
+        // A TORN record propagates (typed) — only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards that have no data
       }
     }
@@ -2283,7 +2323,9 @@ export abstract class BaseStorage extends BaseStorageAdapter {
               batch.map(async (id) => {
                 try {
                   return { id, metadata: await this.getNounMetadata(id) }
-                } catch {
+                } catch (error) {
+                  // A TORN record must surface typed, never as a skipped id.
+                  if (isTornRecordError(error)) throw error
                   return null
                 }
               })
@@ -2305,6 +2347,8 @@ export abstract class BaseStorage extends BaseStorageAdapter {
           }
         }
       } catch (error) {
+        // A TORN record propagates (typed) — only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards with no data
       }
     }
@@ -2515,10 +2559,15 @@ export abstract class BaseStorage extends BaseStorageAdapter {
             // reserved fields top-level, ONLY custom fields in `metadata`.
             collected.push({ verb: this.hydrateVerbWithMetadata(verb, metadata), shard })
           } catch (error) {
+            // A TORN record must surface typed — a paginated read that
+            // silently skips a corrupt row hides data loss from the caller.
+            if (isTornRecordError(error)) throw error
             // Skip verbs that fail to load
           }
         }
       } catch (error) {
+        // A TORN record propagates (typed) — only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards that have no data
       }
     }
@@ -3669,9 +3718,18 @@ export abstract class BaseStorage extends BaseStorageAdapter {
       )
 
       for (const result of chunkResults) {
-        if (result.status === 'fulfilled' && result.value.data !== null) {
+        if (result.status === 'fulfilled') {
+          if (result.value.data !== null) {
             results.set(result.value.path, result.value.data)
           }
+        } else {
+          // A rejected read is a torn record or a real storage fault — NOT an
+          // absent object. Batch hydration backs entity reads (getNounBatch /
+          // getVerbsBatch / find hydration); swallowing the rejection would
+          // silently drop a row the caller cannot distinguish from "never
+          // existed". Propagate the typed/real error loudly instead.
+          throw result.reason
+        }
       }
     }
 
@@ -4636,10 +4694,15 @@ export abstract class BaseStorage extends BaseStorageAdapter {
               }
             }
           } catch (error) {
+            // A TORN record must surface typed — an enumeration that silently
+            // skips a corrupt row hides data loss from the caller.
+            if (isTornRecordError(error)) throw error
             // Skip nouns that fail to load
           }
         }
       } catch (error) {
+        // A TORN record propagates (typed) — only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards that have no data
       }
     }
@@ -4825,11 +4888,16 @@ export abstract class BaseStorage extends BaseStorageAdapter {
               results.push(this.hydrateVerbWithMetadata(verb, metadata))
             }
           } catch (error) {
+            // A TORN record must surface typed — an enumeration that silently
+            // skips a corrupt row hides data loss from the caller.
+            if (isTornRecordError(error)) throw error
             // Skip verbs that fail to load
             prodLog.debug(`[BaseStorage] Failed to load verb from ${verbPath}:`, error)
           }
         }
       } catch (error) {
+        // A TORN record propagates (typed) — only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards that have no data
       }
     }
@@ -4945,6 +5013,9 @@ export abstract class BaseStorage extends BaseStorageAdapter {
           sourceVerbs.push(hydratedVerb)
         }
       } catch (error) {
+        // A TORN record propagates (typed) — batch hydration must not
+        // silently drop a corrupt row. Only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards that have no data
       }
     }
@@ -5030,10 +5101,15 @@ export abstract class BaseStorage extends BaseStorageAdapter {
               results.push(this.hydrateVerbWithMetadata(verb, metadata))
             }
           } catch (error) {
+            // A TORN record must surface typed — an enumeration that silently
+            // skips a corrupt row hides data loss from the caller.
+            if (isTornRecordError(error)) throw error
             // Skip verbs that fail to load
           }
         }
       } catch (error) {
+        // A TORN record propagates (typed) — only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards that have no data
       }
     }
@@ -5078,10 +5154,15 @@ export abstract class BaseStorage extends BaseStorageAdapter {
               )
             )
           } catch (error) {
+            // A TORN record must surface typed — an enumeration that silently
+            // skips a corrupt row hides data loss from the caller.
+            if (isTornRecordError(error)) throw error
             // Skip verbs that fail to load
           }
         }
       } catch (error) {
+        // A TORN record propagates (typed) — only shard-listing absence is skippable.
+        if (isTornRecordError(error)) throw error
         // Skip shards that have no data
       }
     }

src/storage/tornRecordError.ts

@@ -0,0 +1,132 @@
+/**
+ * @module storage/tornRecordError
+ * @description Typed surface for TORN records — files that EXIST in storage but
+ * cannot be decoded (invalid JSON, truncated/garbled gzip). A torn record is
+ * disk corruption, not absence: reading it as `null` ("not found") makes the
+ * consumer unable to distinguish "never existed" from "exists but unreadable",
+ * so nothing ever heals it. Mandate: loud errors, never quiet losses.
+ *
+ * Contract implemented across the storage layer:
+ * - Genuine absence (ENOENT) still reads as clean `null` — no error, no noise.
+ * - A torn record ALWAYS registers here (error log + per-process gauge), then:
+ *   - entity read paths (get/getBatch/pagination/enumeration hydration) throw
+ *     {@link TornRecordError} to the caller — a row is never silently dropped;
+ *   - system-artifact read paths whose machinery is designed for
+ *     absent-artifact degradation (manifests with recovery paths, markers
+ *     whose verdict is "rescan", rebuildable statistics) map torn → their
+ *     existing degrade AFTER the encounter is logged and counted.
+ */
+
+import { prodLog } from '../utils/logger.js'
+
+/**
+ * @description Thrown when a stored object EXISTS but cannot be decoded —
+ * corrupt/torn bytes on disk (invalid JSON, undecodable gzip). Deliberately
+ * distinct from absence: `readObjectFromPath` returns `null` only for ENOENT.
+ * Catchable by type (`instanceof`), by `name === 'TornRecordError'`, or by
+ * `code === 'TORN_RECORD'` (cross-realm safe; never matches `isAbsentError`).
+ */
+export class TornRecordError extends Error {
+  /** Stable machine-checkable discriminator (errno-style). */
+  public readonly code = 'TORN_RECORD'
+  /** Storage-root-relative path of the torn object. */
+  public readonly path: string
+  /** The underlying decode failure (SyntaxError, zlib error, …). */
+  public override readonly cause: unknown
+
+  /**
+   * @param path - Storage-root-relative path of the torn object.
+   * @param cause - The underlying decode failure.
+   */
+  constructor(path: string, cause: unknown) {
+    const causeMessage =
+      cause instanceof Error ? cause.message : String(cause)
+    super(
+      `Torn record at '${path}': file exists but cannot be decoded (${causeMessage}). ` +
+        `This is storage corruption, not absence — the record was not silently skipped.`
+    )
+    this.name = 'TornRecordError'
+    this.path = path
+    this.cause = cause
+  }
+}
+
+/**
+ * @description True IFF `e` is a torn-record error — matches by `instanceof`
+ * first, then by `name`/`code` so errors crossing module-duplication or realm
+ * boundaries are still recognized.
+ * @param e - The caught value.
+ * @returns Whether `e` denotes an existing-but-undecodable stored object.
+ */
+export function isTornRecordError(e: unknown): e is TornRecordError {
+  if (e instanceof TornRecordError) return true
+  if (e === null || typeof e !== 'object') return false
+  const { name, code } = e as { name?: unknown; code?: unknown }
+  return name === 'TornRecordError' || code === 'TORN_RECORD'
+}
+
+/**
+ * @description True IFF `e` is a payload-decode failure — the file's BYTES were
+ * read fine but could not be turned back into an object: `SyntaxError` from
+ * `JSON.parse`, or a zlib error (`Z_DATA_ERROR`, `Z_BUF_ERROR`, …) from gunzip.
+ * Distinguishes "torn record" from real I/O faults (EIO/EACCES/…), which must
+ * propagate as themselves.
+ * @param e - The caught value.
+ * @returns Whether the error means "bytes present, content undecodable".
+ */
+export function isUnparseablePayloadError(e: unknown): boolean {
+  if (e === null || typeof e !== 'object') return false
+  if (e instanceof SyntaxError) return true
+  const { name, code } = e as { name?: unknown; code?: unknown }
+  if (name === 'SyntaxError') return true
+  return typeof code === 'string' && code.startsWith('Z_')
+}
+
+/** Per-process torn-record gauge state (module-scoped; see the accessors). */
+let tornRecordCount = 0
+let lastTornRecordPath: string | null = null
+
+/**
+ * @description Register a torn-record encounter: logs a production ERROR
+ * naming the path, increments the per-process gauge, and returns the typed
+ * error for the caller to throw (or to map into a documented loud degrade).
+ * EVERY torn encounter goes through here, whatever the caller decides —
+ * the floor is: never silent.
+ * @param path - Storage-root-relative path of the torn object.
+ * @param cause - The underlying decode failure.
+ * @returns The constructed {@link TornRecordError}.
+ */
+export function registerTornRecordEncounter(
+  path: string,
+  cause: unknown
+): TornRecordError {
+  tornRecordCount++
+  lastTornRecordPath = path
+  const error = new TornRecordError(path, cause)
+  prodLog.error(
+    `[Storage] TORN RECORD #${tornRecordCount}: '${path}' exists but cannot be decoded — ` +
+      `corrupt or partially written bytes. Cause: ${
+        cause instanceof Error ? `${cause.name}: ${cause.message}` : String(cause)
+      }`
+  )
+  return error
+}
+
+/**
+ * @description Read the per-process torn-record gauge: how many torn records
+ * this process has encountered and the most recent path. Observability seam —
+ * lets operators and tests confirm that corruption was seen, not swallowed.
+ * @returns The current gauge snapshot.
+ */
+export function getTornRecordGauge(): { count: number; lastPath: string | null } {
+  return { count: tornRecordCount, lastPath: lastTornRecordPath }
+}
+
+/**
+ * @description Reset the per-process torn-record gauge to zero. Test seam only
+ * (the gauge is process-lifetime state); production code never resets it.
+ */
+export function resetTornRecordGauge(): void {
+  tornRecordCount = 0
+  lastTornRecordPath = null
+}

src/types/brainy.types.ts

@@ -2084,6 +2084,28 @@ export interface BrainyConfig {
    * `'manual'` restores the pre-9.1 behavior: the engine never flushes on
    * its own (except at `close()`); the caller owns the cadence.
    */
+  /**
+   * Storage-authority posture at open (10.0.0+ fleet default: `'adopt'`).
+   *
+   * `'adopt'` — a brain with NO stored authority artifact adopts LOG
+   * AUTHORITY at open, oracle-gated: the verification oracle replays the
+   * generation log against stored truth; curable divergences (pre-log
+   * rows, witness drift) are baseline-backfilled; the brain flips ONLY on
+   * a green verdict and writes the durable per-brain switch. On green,
+   * writes become durable-at-ack (group-committed log fsync covers every
+   * ack). A brain whose oracle cannot go green STAYS tree-authoritative,
+   * says so loudly, and records the refusal — never a silent half-state.
+   *
+   * `'defer'` — the explicit opt-out: no automatic adoption; the brain
+   * stays tree-authoritative until `adoptLogAuthority()` is called. The
+   * pre-10 behavior, documented for operators who stage their own flips.
+   *
+   * A STORED artifact always wins over this setting (checked-at-open law):
+   * an already-flipped brain stays flipped; an explicitly-recorded tree
+   * posture is honored until an operator re-runs adoption.
+   */
+  logAuthority?: 'adopt' | 'defer'
+
   persistence?: {
     policy?: 'auto' | 'manual'
     /** Background flush after this many committed writes (default 512). */

tests/helpers/durabilityKillMatrix.ts

@@ -60,15 +60,25 @@ export function makeTempDir(): string {
  * Open a writer brain over `dir` with every implicit durability knob off:
  * persistence policy 'manual' (the engine never flushes on its own, so every
  * durable transition in a test is an explicit `flush()`/commit), deterministic
- * embeddings (tests always pass explicit vectors anyway), silent logs.
+ * embeddings (tests always pass explicit vectors anyway), silent logs — and
+ * `logAuthority: 'defer'` (the explicit opt-out of the 10.0.0 adopt-at-open
+ * fleet default), so the durability POSTURE is explicit per row too: rows
+ * pinning deferred/tree recovery semantics get exactly that, and at-ack rows
+ * engage log authority via `flipToAtAck`. The fleet default's open-time
+ * adoption would inject a baseline-backfill generation into every floor
+ * computation and pre-flip every row.
  */
-export async function openBrain(dir: string): Promise<Brainy> {
+export async function openBrain(
+  dir: string,
+  opts?: { logAuthority?: 'adopt' | 'defer' }
+): Promise<Brainy> {
   process.env.BRAINY_DETERMINISTIC_EMBEDDINGS = 'true'
   const brain = new Brainy({
     requireSubtype: false,
     storage: { type: 'filesystem', path: dir },
     silent: true,
-    persistence: { policy: 'manual' }
+    persistence: { policy: 'manual' },
+    logAuthority: opts?.logAuthority ?? 'defer'
   })
   await brain.init()
   return brain

tests/integration/db-mvcc.test.ts

@@ -96,11 +96,15 @@ describe('8.0 Db API — generational MVCC', () => {
   }
 
   /** Open (and track) a filesystem brain rooted at a fresh temp directory. */
-  async function openFsBrain(dir?: string): Promise<{ brain: Brainy; dir: string }> {
+  async function openFsBrain(
+    dir?: string,
+    logAuthority?: 'adopt' | 'defer'
+  ): Promise<{ brain: Brainy; dir: string }> {
     const rootDirectory = dir ?? makeTempDir()
     const brain = new Brainy({
       requireSubtype: false,
-      storage: { type: 'filesystem', path: rootDirectory }
+      storage: { type: 'filesystem', path: rootDirectory },
+      ...(logAuthority ? { logAuthority } : {})
     })
     await brain.init()
     brains.push(brain)
@@ -647,7 +651,13 @@ describe('8.0 Db API — generational MVCC', () => {
   // ==========================================================================
   it('proof 8 — a crash before the manifest rename recovers to the exact pre-transaction state', async () => {
     const dir = makeTempDir()
-    const { brain: first } = await openFsBrain(dir)
+    // 'defer' (tree authority): this proof pins the TREE commit-point
+    // contract — the manifest rename is the commit, so a crash before it
+    // rolls back. Under the adopt-at-open default (log authority) the same
+    // crash point legitimately REPLAYS the fsynced fact at reopen and the
+    // transaction lands — that contract is pinned in the durability kill
+    // matrix's at-ack rows, not here.
+    const { brain: first } = await openFsBrain(dir, 'defer')
 
     await first.transact([
       {
@@ -689,9 +699,10 @@ describe('8.0 Db API — generational MVCC', () => {
     // the realistic worst case for the recovery path.
     await first.close()
 
-    // Reopen: recovery rolls the uncommitted generation back and rebuilds
-    // the indexes from the repaired records.
-    const { brain: second } = await openFsBrain(dir)
+    // Reopen ('defer' again — a reopen under the adopt default would adopt
+    // and change the recovery path): recovery rolls the uncommitted
+    // generation back and rebuilds the indexes from the repaired records.
+    const { brain: second } = await openFsBrain(dir, 'defer')
     const recovered = await second.get(uid('crash-e'))
     expect((recovered?.metadata as { v: number }).v).toBe(1)
     expect(await second.get(uid('crash-new'))).toBeNull()
@@ -1162,13 +1173,16 @@ describe('8.0 Db API — generational MVCC', () => {
     const brain = await openMemoryBrain()
 
     // Model-B: a single-op write is its OWN generation and IS logged (no meta —
-    // tx metadata is a transact()-only concept). It is generation 1 on a fresh
-    // brain (init-time infrastructure writes are the un-versioned gen-0 baseline).
+    // tx metadata is a transact()-only concept). Relative baseline: under the
+    // adopt-at-open fleet default the open-time baseline backfill is itself a
+    // logged single-op generation, so the log is not empty on a fresh brain —
+    // every pin below is expressed against that baseline.
+    const baseGens = (await brain.transactionLog()).map((entry) => entry.generation)
     await brain.add({ id: uid('txlog-solo'), type: NounType.Document, data: 'solo', vector: vec(99), subtype: 'note' })
     const soloLog = await brain.transactionLog()
-    expect(soloLog.map((entry) => entry.generation)).toEqual([1])
+    const soloGen = brain.generation()
+    expect(soloLog.map((entry) => entry.generation)).toEqual([soloGen, ...baseGens])
     expect(soloLog[0].meta).toBeUndefined()
-    const soloGen = 1
 
     const first = await brain.transact(
       [{ op: 'add', id: uid('txlog-a'), type: NounType.Document, data: 'a', vector: vec(100), metadata: {} }],
@@ -1181,12 +1195,14 @@ describe('8.0 Db API — generational MVCC', () => {
     const third = await brain.transact([{ op: 'update', id: uid('txlog-a'), metadata: { v: 3 } }])
 
     const entries = await brain.transactionLog()
-    // Newest first: the three transacts, then the single-op solo write (gen 1).
+    // Newest first: the three transacts, then the single-op solo write, then
+    // whatever the open baseline logged (the adopt-at-open backfill).
     expect(entries.map((entry) => entry.generation)).toEqual([
       third.generation,
       second.generation,
       first.generation,
-      soloGen
+      soloGen,
+      ...baseGens
     ])
     expect(entries[1].meta).toEqual({ author: 'job-2' })
     expect(entries[2].meta).toEqual({ author: 'job-1' })
@@ -1238,21 +1254,24 @@ describe('8.0 Db API — generational MVCC', () => {
     const brain = await openMemoryBrain()
     const a = uid('ov-a')
     const b = uid('ov-b')
-    await (
-      await brain.transact([
+    // Pin RELATIVELY at the transact's own generation (not an absolute 1 —
+    // the adopt-at-open baseline backfill owns the first generation).
+    const tx = await brain.transact([
       { op: 'add', id: a, type: NounType.Document, data: 'a', vector: vec(1), metadata: { v: 1 } },
       { op: 'add', id: b, type: NounType.Document, data: 'b', vector: vec(2), metadata: { v: 1 } }
     ])
-    ).release()
-    const at1 = await brain.asOf(1)
+    const txGen = tx.generation
+    await tx.release()
+    const at1 = await brain.asOf(txGen)
 
     // A single-op REMOVE of `b` lands AFTER the pin and is NOT flushed (pending).
     await brain.remove(b)
 
     const liveIds = (await brain.find({})).map((r) => r.id)
     const pastIds = (await at1.find({})).map((r) => r.id)
-    // Live: `b` is gone. Historical (pinned at gen 1): the un-flushed removal is
-    // overlaid out, so `b` is still present at its pinned state.
+    // Live: `b` is gone. Historical (pinned at the transact's generation): the
+    // un-flushed removal is overlaid out, so `b` is still present at its
+    // pinned state.
     expect(liveIds).toContain(a)
     expect(liveIds).not.toContain(b)
     expect(pastIds).toContain(a)
@@ -1262,11 +1281,14 @@ describe('8.0 Db API — generational MVCC', () => {
 
   it('Model-B retention — explicit caps reclaim single-op history; committed history survives reopen', async () => {
     const { brain, dir } = await openFsBrain()
+    // Relative baseline: the adopt-at-open backfill holds the first
+    // generation(s), so the 6 writes below land at base+1..base+6.
+    const base = brain.generation()
     const a = uid('ret-a')
     await brain.add({ id: a, type: NounType.Document, data: 'a', vector: vec(1), metadata: { v: 1 } })
     for (let v = 2; v <= 6; v++) await brain.update({ id: a, metadata: { v } })
     await brain.flush() // persist the per-write generations to disk
-    expect(brain.generation()).toBe(6)
+    expect(brain.generation()).toBe(base + 6)
 
     // Cap to the 2 most recent generations — older single-op history is reclaimed.
     const res = await brain.compactHistory({ maxGenerations: 2 })

tests/integration/db-temporal.test.ts

@@ -36,6 +36,9 @@ import { GenerationCompactedError } from '../../src/db/errors.js'
 import type { GenerationStore } from '../../src/db/generationStore.js'
 import { NounType } from '../../src/types/graphTypes.js'
 
+/** The VFS root — re-committed by the adopt-at-open baseline backfill. */
+const VFS_ROOT = '00000000-0000-0000-0000-000000000000'
+
 /** Deterministic 384-dim vector so no test ever invokes the embedder. */
 function vec(seed: number): number[] {
   return Array.from({ length: 384 }, (_, i) => ((seed * 31 + i * 7) % 100) / 100)
@@ -133,7 +136,11 @@ describe('8.0 Db API — temporal range verbs', () => {
     expect(viaDb).toEqual(viaGen)
     expect(viaDb.fromGeneration).toBe(g1)
     expect(viaDb.nouns).toEqual([a, b].sort()) // a (updated after g1) + b (added after g1)
-    expect(viaEpoch.nouns).toEqual([a, b].sort()) // (0, now] also includes a's creation, still {a, b}
+    // (0, now] also includes a's creation — still {a, b} among user rows. The
+    // adopt-at-open baseline backfill re-commits the VFS root as a real
+    // generation, so the full-epoch window legitimately reports it too;
+    // filter it to keep this pin about the user writes.
+    expect(viaEpoch.nouns.filter((n) => n !== VFS_ROOT)).toEqual([a, b].sort())
 
     // direction guard: an older view cannot be `since` a newer lower bound
     const older = await brain.asOf(1)
@@ -163,7 +170,11 @@ describe('8.0 Db API — temporal range verbs', () => {
     }
 
     const all = await brain.transactionLog()
-    expect(all.map((e) => e.generation)).toEqual([...gens].reverse()) // newest first
+    // Newest first — compared above the open baseline (the adopt-at-open
+    // backfill logs its own generation(s) below the first user write).
+    expect(all.map((e) => e.generation).filter((g) => g >= gens[0])).toEqual(
+      [...gens].reverse()
+    )
 
     // INCLUSIVE both ends — gens[1] AND gens[3] are present (contrast since's exclusive lower).
     const windowed = await brain.transactionLog({ from: gens[1], to: gens[3] })
@@ -334,19 +345,22 @@ describe('8.0 Db API — temporal range verbs', () => {
   // 7. Granularity (Model-B) ---------------------------------------------------
   it('granularity: single-operation writes ARE versioned and visible to the temporal verbs', async () => {
     const brain = await openMemoryBrain()
+    // Relative baseline: the adopt-at-open backfill already logged its own
+    // generation(s) — pin the DELTA this test's writes add, not a count.
+    const baseCount = (await brain.transactionLog()).length
     const a = uid('gran-a')
     const r1 = await brain.transact([
       { op: 'add', id: a, type: NounType.Document, data: 'a', vector: vec(1), metadata: { v: 1 } }
     ])
     await r1.release()
-    expect((await brain.transactionLog()).length).toBe(1)
+    expect((await brain.transactionLog()).length).toBe(baseCount + 1)
 
     // Model-B: a single-op write is its OWN immutable generation — logged,
     // diffable, and time-travelable, exactly like a transact() of one op.
     await brain.update({ id: a, metadata: { v: 2 } })
 
     // The single-op update appended a generation/log entry.
-    expect((await brain.transactionLog()).length).toBe(2)
+    expect((await brain.transactionLog()).length).toBe(baseCount + 2)
     expect(brain.generation()).toBe(r1.generation + 1)
 
     // diff sees the single-op update as a modification of `a`.

tests/integration/durability-kill-matrix.test.ts

@@ -109,14 +109,14 @@ describe('durability kill matrix — crash at every commit-path step, recover by
   /**
    * Flip a brain to durable-at-ack (log-authority) mode.
    *
-   * NOT via `adoptLogAuthority()`: the sanctioned flip REFUSES on a freshly
-   * materialized brain — its verification oracle reports the generation-0
-   * VFS-root baseline as a divergence (`state-differs` even after an
-   * identity-update backfill; verified 2026-08-10). This helper flips the
-   * SAME switch the sanctioned path flips (`setLogDurability('at-ack')`) and
-   * persists the SAME authority artifact, so a reopened brain also runs in
-   * log-authority mode. The durability semantics under test are governed
-   * entirely by that switch.
+   * NOT via `adoptLogAuthority()` (and the helper opens every brain with
+   * `logAuthority: 'defer'`, opting out of the 10.0.0 adopt-at-open fleet
+   * default): the sanctioned path runs the oracle and a baseline backfill,
+   * which appends its own generation — shifting the floor arithmetic every
+   * row pins. This helper flips the SAME switch the sanctioned path flips
+   * (`setLogDurability('at-ack')`) and persists the SAME authority artifact,
+   * so a reopened brain also runs in log-authority mode. The durability
+   * semantics under test are governed entirely by that switch.
    */
   async function flipToAtAck(brain: Brainy): Promise<void> {
     const storage = (

tests/integration/fact-log-contracts.test.ts

@@ -4,14 +4,13 @@
  *
  * (1) FSYNC-BEFORE-ACK: an acknowledged write's fact survives an abrupt
  *     process end (no flush, no close — reopen from disk).
- *     - transact(): HOLDS TODAY — the fact is fsync'd before transact returns.
- *     - single-op: PINNED AS `it.fails` — today's group-commit batches
- *       DURABILITY (ack precedes the group fsync; a hard kill loses the fact
- *       AND the generation together, coherently — the documented Model-B
- *       contract, fine while the tree is authoritative). The destination
- *       (ack-at-log) requires group commit to become LATENCY batching: the
- *       ack waits for the shared fsync. When that lands, this pin flips red —
- *       remove `.fails` and the contract is permanent. No cliff to discover.
+ *     - transact(): HOLDS — the fact is fsync'd before transact returns.
+ *     - single-op: HOLDS (was pinned `it.fails` until the ack-at-log
+ *       destination landed): the 10.0.0 adopt-at-open fleet default flips a
+ *       fresh brain to log authority at open, so single-op acks await the
+ *       covering group fsync (durable-at-ack) and recovery REPLAYS intact
+ *       facts above the manifest at the next open. The contract is now
+ *       permanent on every path.
  *
  * (2) SCAN STABILITY UNDER ROTATION: a scan handle opened before segment
  *     rotation yields exactly its snapshot — byte-identical facts, no gaps,
@@ -63,9 +62,11 @@ describe('fsync-before-ack contract (fact durability at the ack boundary)', () =
     expect(facts.some((f) => f.generation === receipt.generation)).toBe(true)
   })
 
-  // PINNED (flips red when group commit becomes latency batching — then
-  // remove `.fails` and the ack-at-log contract is permanent on every path).
-  it.fails('single-op: the fact is durable the moment the ack returns (the ack-at-log target)', async () => {
+  // THE ACK-AT-LOG CONTRACT, HELD (was `.fails` until it landed): under the
+  // adopt-at-open fleet default this brain runs durable-at-ack from open —
+  // the ack waits for the covering log fsync, and the log-authority recovery
+  // path replays the intact fact at the next open instead of truncating it.
+  it('single-op: the fact is durable the moment the ack returns (the ack-at-log target)', async () => {
     await brain.add({ data: 'acked single-op', type: 'document', metadata: { n: 1 } })
     const ackedHead = brain.scanFacts()!.headGeneration
     // Abrupt end immediately after the ack — before any flush window.

tests/integration/log-authority-adopt.test.ts

@@ -22,8 +22,12 @@ afterEach(async () => {
   for (const d of dirs.splice(0)) rmSync(d, { recursive: true, force: true })
 })
 
-async function open(dir: string): Promise<Brainy> {
-  const b = new Brainy({ storage: { type: 'filesystem', path: dir }, requireSubtype: false })
+async function open(dir: string, logAuthority?: 'adopt' | 'defer'): Promise<Brainy> {
+  const b = new Brainy({
+    storage: { type: 'filesystem', path: dir },
+    requireSubtype: false,
+    ...(logAuthority ? { logAuthority } : {})
+  })
   await b.init()
   brains.push(b)
   return b
@@ -80,4 +84,31 @@ describe('adoptLogAuthority — the sanctioned flip with self-backfill', () => {
     expect(report.verdict).toBe('green')
     expect(brain.logAuthority().authority).toBe('log')
   }, 120000)
+
+  // THE OPT-OUT CONTRACT (`logAuthority: 'defer'`): no automatic adoption —
+  // the fresh brain stays tree-authoritative and writes NO artifact (a
+  // deferred posture is config, not stored state); the EXPLICIT
+  // adoptLogAuthority() then flips it exactly as before the fleet default.
+  it("opt-out: 'defer' stays tree with no artifact until the explicit adoptLogAuthority() flips it", async () => {
+    const dir = mkdtempSync(join(tmpdir(), 'brainy-adopt-defer-'))
+    dirs.push(dir)
+    const brain = await open(dir, 'defer')
+    await brain.add({ data: 'deferred row', type: NounType.Document, metadata: { n: 1 } })
+    await brain.flush()
+
+    expect(brain.logAuthority().authority, "'defer' skips open-time adoption").toBe('tree')
+    const storage = (brain as unknown as {
+      storage: { readRawObject(p: string): Promise<unknown | null> }
+    }).storage
+    const artifact = await storage.readRawObject('_system/log-authority.json').catch(() => null)
+    expect(artifact, "'defer' writes no authority artifact").toBeNull()
+
+    const report = await brain.adoptLogAuthority()
+    expect(report.verdict, 'the explicit flip still lands on green').toBe('green')
+    expect(brain.logAuthority().authority).toBe('log')
+    const stored = (await storage.readRawObject('_system/log-authority.json')) as {
+      authority?: string
+    } | null
+    expect(stored?.authority, 'the explicit flip stores the artifact').toBe('log')
+  }, 120000)
 })

tests/integration/log-authority.test.ts

@@ -1,22 +1,34 @@
 /**
  * @module tests/integration/log-authority
  * @description The guarded log-authority core, end-to-end: the per-brain
- * authority switch (default 'tree', stored artifact, checked at open only),
- * the verification oracle (replay the fact log, diff latest per-id state
+ * authority switch (stored artifact, checked at open only), the
+ * verification oracle (replay the fact log, diff latest per-id state
  * against the canonical tree, NAME every divergence by class), the guarded
  * flip (refuses on red with the cure in the message; lands on green and
  * engages durable-at-ack immediately), and the switch surviving reopen.
  *
+ * THE 10.0.0 FLEET DEFAULT is ADOPT-AT-OPEN (`logAuthority: 'adopt'`): a
+ * fresh brain with no stored artifact runs the oracle at open, backfills
+ * curable divergences, and flips to log authority on green — so a
+ * default-config brain opens ALREADY log-authoritative and durable-at-ack.
+ * The first two pins hold that default and its explicit opt-out
+ * (`logAuthority: 'defer'`, the pre-10 tree behavior). Every test below
+ * them that exercises the ORACLE or the EXPLICIT flip opens its brain with
+ * `'defer'` — otherwise the open-time adoption would have pre-flipped the
+ * brain and pre-cured the very divergences under test.
+ *
  * KNOWN GAPS PINNED WITH `.fails` (real findings, not test bugs — see the
  * comments on each): a fresh brain is NOT log-complete by construction
  * today, because the VFS root is written at init as a baseline
  * (generation-less) write that never gets a fact, so the oracle reports it
- * as a `pre-log-record` and no fresh brain can flip without a manual
- * baseline backfill. The tests that need a green oracle perform that
- * backfill explicitly (an identity update of the root as the FINAL write —
- * final, because derived-index maintenance rewrites canonical noun records
- * outside generations, so an earlier fact's after-image goes stale; see the
- * module tail comment on `backfillBaseline`).
+ * as a `pre-log-record`. The open-time adoption (and adoptLogAuthority())
+ * CURES this by baseline backfill — a re-commit, not construction — so the
+ * by-construction pin stays `.fails` on a deferred brain. Tests that need
+ * a green oracle on a deferred brain perform that backfill explicitly (an
+ * identity update of the root as the FINAL write — final, because
+ * derived-index maintenance rewrites canonical noun records outside
+ * generations, so an earlier fact's after-image goes stale; see the module
+ * tail comment on `backfillBaseline`).
  */
 import { describe, it, expect, afterEach } from 'vitest'
 import { mkdtempSync, rmSync } from 'node:fs'
@@ -88,14 +100,24 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   const dirs: string[] = []
   const brains: Brainy[] = []
 
-  const openBrain = async (dir?: string): Promise<{ brain: Brainy; dir: string }> => {
+  /**
+   * Open a brain over `dir`. Omit `logAuthority` to exercise the FLEET
+   * DEFAULT (adopt-at-open); pass `'defer'` for the tests that need a
+   * tree-authoritative brain so the oracle/explicit-flip path is actually
+   * the thing under test (the default would pre-flip and pre-backfill).
+   */
+  const openBrain = async (
+    dir?: string,
+    logAuthority?: 'adopt' | 'defer'
+  ): Promise<{ brain: Brainy; dir: string }> => {
     const d = dir ?? mkdtempSync(join(tmpdir(), 'brainy-log-authority-'))
     if (!dir) dirs.push(d)
     const brain = new Brainy({
       storage: { type: 'filesystem', path: d },
       requireSubtype: false,
       silent: true,
-      dimensions: 384
+      dimensions: 384,
+      ...(logAuthority ? { logAuthority } : {})
     })
     brains.push(brain)
     await brain.init()
@@ -109,8 +131,37 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
     for (const d of dirs.splice(0)) rmSync(d, { recursive: true, force: true })
   })
 
-  it('DEFAULT IS TREE: a fresh brain reports tree authority, stores no artifact, and plain acks never await a log fsync', async () => {
-    const { brain } = await openBrain()
+  // THE RULED DEFAULT (10.0.0): with no config and no stored artifact, a
+  // fresh brain ADOPTS log authority at open — oracle green (the open-time
+  // baseline backfill cures the generation-0 VFS root), artifact on disk,
+  // durable-at-ack live from the first write.
+  it('DEFAULT IS ADOPT-AT-OPEN: a fresh brain opens already log-authoritative — artifact stored, plain acks await the covering log fsync', async () => {
+    const { brain } = await openBrain() // no logAuthority config = the fleet default
+
+    const authority = brain.logAuthority()
+    expect(authority.authority).toBe('log')
+    expect(typeof authority.flippedAt).toBe('number')
+    expect(authority.oracle, 'the open-time flip records its green oracle summary').toBeDefined()
+
+    const artifact = (await internals(brain)
+      .storage.readRawObject(AUTHORITY_ARTIFACT)
+      .catch(() => null)) as { authority?: string } | null
+    expect(artifact, 'the adoption wrote the switch artifact').not.toBeNull()
+    expect(artifact!.authority).toBe('log')
+
+    // The MODE assertion (not a timing one): in log authority a single-op
+    // ack awaits the log's covering-fsync path.
+    expect(internals(brain).generationStore.logDurability).toBe('at-ack')
+    const spy = spyEnsureSynced(brain)
+    await brain.add({ data: 'log mode write', type: 'document', metadata: { n: 1 } })
+    expect(spy.calls(), 'adopted default: add() awaits the covering fsync').toBeGreaterThanOrEqual(1)
+  })
+
+  // THE EXPLICIT OPT-OUT: `logAuthority: 'defer'` is the pre-10 behavior —
+  // tree authority, NO artifact written (a deferred posture is config, not
+  // stored state), and single-op acks never await a log fsync.
+  it("OPT-OUT ('defer'): the brain stays tree-authoritative, stores no artifact, and plain acks never await a log fsync", async () => {
+    const { brain } = await openBrain(undefined, 'defer')
 
     expect(brain.logAuthority().authority).toBe('tree')
     expect(brain.logAuthority().flippedAt).toBeUndefined()
@@ -118,7 +169,7 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
     const artifact = await internals(brain)
       .storage.readRawObject(AUTHORITY_ARTIFACT)
       .catch(() => null)
-    expect(artifact, 'no switch artifact exists before any flip').toBeNull()
+    expect(artifact, "'defer' writes no switch artifact").toBeNull()
 
     // The MODE assertion (not a timing one): in tree authority a single-op
     // ack must never call the log's covering-fsync path.
@@ -134,10 +185,12 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   // (00000000-0000-0000-0000-000000000000) is created at init by a baseline
   // write with NO generation and NO fact, yet it is enumerated by the
   // canonical walk — so the oracle on a fresh brain is red with exactly one
-  // `pre-log-record` mismatch on the root, and adoptLogAuthority() refuses
-  // on every fresh brain. Verified empirically on this branch.
+  // `pre-log-record` mismatch on the root. The adopt-at-open default (and
+  // adoptLogAuthority()) CURES this by baseline backfill — a re-commit,
+  // which is why this pin opens with 'defer': it holds the BY-CONSTRUCTION
+  // intent, which the backfill masks but does not deliver.
   it.fails('ORACLE INTENT: a fresh brain is log-complete by construction — verdict green with zero mismatches', async () => {
-    const { brain } = await openBrain()
+    const { brain } = await openBrain(undefined, 'defer')
     await seedWrites(brain)
     await brain.flush()
 
@@ -147,7 +200,9 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   })
 
   it('a fresh, un-backfilled brain diverges ONLY on the init-time baseline record — every user write is exactly reproduced', async () => {
-    const { brain } = await openBrain()
+    // 'defer': the adopt-at-open default would have backfilled the baseline
+    // already — this pin needs the brain genuinely un-backfilled.
+    const { brain } = await openBrain(undefined, 'defer')
     await seedWrites(brain)
     await brain.flush()
 
@@ -166,7 +221,10 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   })
 
   it('THE ORACLE GOES GREEN on a log-complete brain: adds + update + remove, every canonical row exactly reproduced', async () => {
-    const { brain } = await openBrain()
+    // 'defer' + manual backfill: the exact-count pins below (5 generations)
+    // depend on the log holding ONLY this test's writes — the adopt-at-open
+    // default would inject its own backfill generation at init.
+    const { brain } = await openBrain(undefined, 'defer')
     await seedWrites(brain)
     await backfillBaseline(brain) // final write — see the helper's contract
     await brain.flush()
@@ -184,7 +242,7 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   })
 
   it('THE ORACLE NAMES pre-log records: a canonical row no fact ever recorded reports pre-log-record, by id', async () => {
-    const { brain } = await openBrain()
+    const { brain } = await openBrain(undefined, 'defer')
     await seedWrites(brain)
     await backfillBaseline(brain)
     await brain.flush()
@@ -226,7 +284,9 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
     // and the flip proceeds; ONLY log-AHEAD divergences (the log claims
     // state canonical denies) refuse, because no backfill can make the log
     // un-claim a live row. This test stages exactly that incurable shape.
-    const { brain } = await openBrain()
+    // 'defer': the brain must still be tree-authoritative (no artifact) so
+    // the refusal's nothing-written pins below have meaning.
+    const { brain } = await openBrain(undefined, 'defer')
     const { kept } = await seedWrites(brain)
     await backfillBaseline(brain)
     await brain.flush()
@@ -254,7 +314,9 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   })
 
   it('THE FLIP LANDS ON GREEN: the report is the receipt, the artifact is on disk, and durable-at-ack engages immediately', async () => {
-    const { brain } = await openBrain()
+    // 'defer': this pin exercises the EXPLICIT flip — the adopt-at-open
+    // default would have landed it before the test began.
+    const { brain } = await openBrain(undefined, 'defer')
     await seedWrites(brain)
     await backfillBaseline(brain)
     await brain.flush()
@@ -284,7 +346,7 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   })
 
   it('THE SWITCH SURVIVES REOPEN: authority restored at open with no re-verification, durable-at-ack active in the new session', async () => {
-    const { brain, dir } = await openBrain()
+    const { brain, dir } = await openBrain(undefined, 'defer')
     await seedWrites(brain)
     await backfillBaseline(brain)
     await brain.flush()
@@ -292,7 +354,10 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
     const flipReceipt = brain.logAuthority()
     await (brain as unknown as { close: () => Promise<void> }).close()
 
-    const { brain: reopened } = await openBrain(dir)
+    // Reopen with 'defer' too: the restored authority below can then ONLY
+    // come from the stored artifact (a stored artifact always wins; had the
+    // default re-adopted, flippedAt/oracle would differ from the receipt).
+    const { brain: reopened } = await openBrain(dir, 'defer')
     const restored = reopened.logAuthority()
     expect(restored.authority).toBe('log')
     // No re-verification happened at open: the restored record IS the stored
@@ -308,7 +373,7 @@ describe('log authority — the switch, the oracle, the guarded flip', () => {
   })
 
   it('STATE-DIFFERS: canonical drift the write path never saw is named, by id', async () => {
-    const { brain } = await openBrain()
+    const { brain } = await openBrain(undefined, 'defer')
     const { kept } = await seedWrites(brain)
     await backfillBaseline(brain)
     await brain.flush()

tests/integration/transact-durability-barrier.test.ts

@@ -43,6 +43,13 @@ describe('transact durability barrier — entity writes fsync before the counter
     })
     await brain.init()
 
+    // Drain the pending tier BEFORE instrumenting: the adopt-at-open fleet
+    // default re-commits the init-time baseline as a buffered single-op
+    // generation, and transact() flushes buffered single-ops first — that
+    // flush's manifest sync would otherwise be recorded ahead of the
+    // transact's own commit point and break the first-index ordering pins.
+    await brain.flush()
+
     // Instrument the real filesystem storage: record every fsync batch in order,
     // and count barrier open/flush, delegating to the originals.
     syncCalls = []

tests/unit/db/bounded-chains.test.ts

@@ -477,14 +477,17 @@ describe('materializeAtGeneration — bounded & deadlock-free (GA #33)', () => {
     const store = (brain as any).generationStore
 
     const N = 400
+    // Relative, not absolute: under the adopt-at-open default the open-time
+    // baseline backfill takes a generation of its own, so the first add is
+    // NOT generation 1 — pin the deep generation to the first add's commit.
+    let deepGen = 0
     for (let i = 0; i < N; i++) {
       await brain.add({ data: `doc ${i}`, type: NounType.Document, subtype: 'note', metadata: { i }, vector: VEC })
+      if (i === 0) deepGen = brain.generation()
     }
     const R = brain.generation() // ≈ N (each add is its own generation)
     expect(R).toBeGreaterThanOrEqual(N)
 
-    const deepGen = 1
-
     // Count getDelta invocations during the materialize.
     const realGetDelta = store.getDelta.bind(store)
     let getDeltaCalls = 0
@@ -509,7 +512,8 @@ describe('materializeAtGeneration — bounded & deadlock-free (GA #33)', () => {
     expect(getDeltaCalls).toBeLessThan(R * 5)
     expect(getDeltaCalls).toBeLessThan(N * N) // the regression guard
 
-    // The materialized at-gen-1 brain holds exactly the one entity that existed.
+    // The materialized brain at the first add's generation holds exactly the
+    // one user entity that existed.
     const atGen1 = await handle.find({ limit: N + 10 })
     expect(atGen1.length).toBe(1)
     await handle.close()

tests/unit/db/fact-log-group-sync.test.ts

@@ -7,12 +7,13 @@
  * one), a solo writer syncs immediately, and at the brain level an at-ack
  * ack resolving means the write's fact is on disk.
  *
- * One pin is marked `.fails` (real finding, not a test bug): the at-ack
- * durability contract says an acked write's fact survives power loss, but
- * FactLog.open() truncates every fact beyond the store's committed
- * generation watermark — which only advances at the pending-tier flush. A
- * crash-shaped reopen (acks landed, flush never ran) therefore DISCARDS the
- * fsynced facts at open. See the test comment for the exact mechanism.
+ * The final pin holds the at-ack durability contract END TO END: an acked
+ * write's fact survives a crash-shaped reopen. This was a `.fails` known
+ * gap (FactLog.open() truncated every fact beyond the committed watermark,
+ * which only advances at the pending-tier flush) — CURED by the 10.0.0
+ * adopt-at-open fleet default: a fresh brain stores the log-authority
+ * artifact at open, and under 'log' authority recovery REPLAYS intact
+ * facts above the manifest instead of truncating them.
  */
 import { describe, it, expect, beforeEach, afterEach } from 'vitest'
 import { mkdtempSync, rmSync } from 'node:fs'
@@ -188,9 +189,9 @@ describe('durable-at-ack through the brain (group commit end-to-end)', () => {
 
   it('at-ack: N concurrent add() acks all resolve, every ack was covered by a log sync, and every fact is on disk after reopen', async () => {
     const { brain, dir } = await openBrain()
-    // White-box: engage the at-ack durability mode directly (the guarded
-    // authority flip that normally enables it is covered by the integration
-    // suite — this test pins the durability machinery itself).
+    // The 10.0.0 fleet default already adopted log authority at open, so
+    // the brain is at-ack; the white-box engage stays so this pin holds the
+    // durability MACHINERY itself independent of the open-time posture.
     brain.generationStore.setLogDurability('at-ack')
 
     const factLog = brain.generationStore.getFactLog()
@@ -231,19 +232,17 @@ describe('durable-at-ack through the brain (group commit end-to-end)', () => {
     }
   })
 
-  // KNOWN GAP (marked .fails — remove the marker when fixed in src): the
-  // at-ack contract is that an acked write's fact survives power loss. The
-  // fsync at ack does put the fact's bytes on disk — but FactLog.open()
-  // truncates every fact with generation > the store's committed watermark,
-  // and that watermark only advances at the pending-tier flush
-  // (flushPendingSingleOps). So on a crash-shaped reopen (acks landed, flush
-  // never ran) the store logs "[FactLog] truncating N uncommitted fact(s)"
-  // and DISCARDS the acked, fsynced facts. Until recovery treats the log as
-  // authoritative past the tree's watermark (or the watermark goes durable
-  // at ack), durable-at-ack does not survive the very crash it exists for.
-  it.fails('at-ack CONTRACT: acked facts survive a crash-shaped reopen (no flush ever ran)', async () => {
+  // THE AT-ACK CONTRACT, HELD (was a `.fails` known gap): an acked write's
+  // fact survives a crash-shaped reopen. Fixed by the 10.0.0 adopt-at-open
+  // fleet default — this brain adopted LOG authority at open (artifact
+  // stored, durable-at-ack live), and under 'log' authority FactLog
+  // recovery REPLAYS intact facts above the committed watermark at the next
+  // open instead of truncating them back. Durable-at-ack now survives the
+  // very crash it exists for.
+  it('at-ack CONTRACT: acked facts survive a crash-shaped reopen (no flush ever ran)', async () => {
     const { brain, dir } = await openBrain()
-    brain.generationStore.setLogDurability('at-ack')
+    expect(brain.logAuthority().authority, 'the fleet default adopted at open').toBe('log')
+    expect(brain.generationStore.logDurability).toBe('at-ack')
     // Crash simulation: the pending-tier durability flush never happens
     // (every trigger routes through flushPendingSingleOps), and the brain is
     // abandoned without close() — exactly the power-loss shape at-ack is for.

tests/unit/db/torn-open-guards.test.ts

@@ -0,0 +1,97 @@
+/**
+ * @module tests/unit/db/torn-open-guards
+ * @description Power-cut throw-site cures (brainy-alone fault-injection
+ * findings, both release-gating):
+ * 1. A torn generation manifest/counter (NaN/garbage where a generation
+ *    belongs) DISCARDS with narration and re-derives — never a RangeError
+ *    killing the open.
+ * 2. A manifest-listed-but-unloadable column segment QUARANTINES at
+ *    discovery with narration; the field serves its remaining segments
+ *    DEGRADED — never a raw throw killing every query on the field.
+ */
+import { describe, it, expect, afterEach } from 'vitest'
+import { mkdtempSync, rmSync, readdirSync, writeFileSync, readFileSync, existsSync } from 'node:fs'
+import { tmpdir } from 'node:os'
+import { join } from 'node:path'
+import { gzipSync } from 'node:zlib'
+import { Brainy } from '../../../src/index.js'
+import { NounType } from '../../../src/types/graphTypes.js'
+
+const dirs: string[] = []
+const brains: Brainy[] = []
+afterEach(async () => {
+  for (const b of brains.splice(0)) await b.close().catch(() => {})
+  for (const d of dirs.splice(0)) rmSync(d, { recursive: true, force: true })
+})
+
+async function open(dir: string): Promise<Brainy> {
+  const b = new Brainy({ storage: { type: 'filesystem', path: dir }, requireSubtype: false })
+  await b.init()
+  brains.push(b)
+  return b
+}
+
+describe('torn-open guards', () => {
+  it('a torn generation manifest (NaN) opens with narrated discard — never a RangeError', async () => {
+    const dir = mkdtempSync(join(tmpdir(), 'brainy-torn-gen-'))
+    dirs.push(dir)
+    let brain = await open(dir)
+    const id = await brain.add({ data: 'survivor row', type: NounType.Document, metadata: { k: 1 } })
+    await brain.flush()
+    await brain.close()
+    brains.pop()
+
+    // The power-cut shape: the manifest's generation field is garbage.
+    const sys = join(dir, '_system')
+    const manifestPath = ['manifest.json', 'manifest.json.gz']
+      .map((f) => join(sys, f))
+      .find((p) => existsSync(p))!
+    const torn = { version: 1, generation: 'NaN-garbage', committedAt: 'x', horizon: null }
+    if (manifestPath.endsWith('.gz')) writeFileSync(manifestPath, gzipSync(JSON.stringify(torn)))
+    else writeFileSync(manifestPath, JSON.stringify(torn))
+
+    // Open MUST succeed (narrated discard + recovery re-derivation), and the
+    // durable row must still serve (log-authority replay recovers it).
+    brain = await open(dir)
+    expect((await brain.get(id))!.data).toContain('survivor row')
+    // Writes continue with a sane monotonic generation.
+    await brain.add({ data: 'post-recovery', type: NounType.Document, metadata: { k: 2 } })
+    expect(Number.isSafeInteger(brain.generation())).toBe(true)
+  }, 120000)
+
+  it('a torn column segment quarantines at discovery; the field serves remaining segments degraded — never a raw throw', async () => {
+    const dir = mkdtempSync(join(tmpdir(), 'brainy-torn-seg-'))
+    dirs.push(dir)
+    let brain = await open(dir)
+    for (let i = 0; i < 6; i++) {
+      await brain.add({ data: `row ${i}`, type: NounType.Document, metadata: { bucket: i % 2 } })
+    }
+    await brain.flush()
+    await brain.close()
+    brains.pop()
+
+    // Tear ONE column segment's bytes on disk (manifest keeps listing it) —
+    // the QUERIED field's own segment, so the quarantine path provably
+    // engages. Column segments live under the raw-blob root:
+    // `<root>/_blobs/_column_index/<field>/L<level>-<id>.bin`.
+    const segDir = join(dir, '_blobs', '_column_index', 'bucket')
+    let tornOne = false
+    if (existsSync(segDir)) {
+      for (const f of readdirSync(segDir, { withFileTypes: true })) {
+        if (!f.isDirectory() && /^L\d+-.*\.bin$/.test(f.name)) {
+          writeFileSync(join(segDir, f.name), Buffer.from([0x00, 0x01, 0x02])) // garbage
+          tornOne = true
+          break
+        }
+      }
+    }
+    expect(tornOne, 'found a segment file to tear (layout probe)').toBe(true)
+
+    // Queries on the field MUST NOT throw — degraded-announced service.
+    brain = await open(dir)
+    const rows = await brain.find({ where: { bucket: 0 }, limit: 10 })
+    expect(Array.isArray(rows), 'query survives the torn segment').toBe(true)
+    // Full completeness is NOT asserted (the torn segment's rows may be
+    // absent — that is the documented degraded contract until heal).
+  }, 120000)
+})

tests/unit/indexes/columnStore/segment-load-fault.test.ts

@@ -5,19 +5,22 @@
  * doing so dropped every entity in that segment out of `filter`/`rangeQuery`/
  * `sortTopK` with no error, so a corrupt index looked like a merely short result.
  *
- * The three failure classes and their required behaviour:
+ * The three failure classes and their required behaviour (torn-segment
+ * QUARANTINE contract — a raw throw at query time killed every query on the
+ * field forever; a silent skip hid the loss; quarantine is the middle):
  *  - a real storage IO fault (EIO) PROPAGATES verbatim — a present-but-unreadable
  *    segment is not "absent", so it must not read as an empty result;
- *  - a manifest-listed segment with undecodable bytes throws `ColumnSegmentLoadError`;
- *  - a manifest-listed segment with NO bytes (gone on disk) throws `ColumnSegmentLoadError`.
+ *  - a manifest-listed segment with undecodable bytes is QUARANTINED at
+ *    discovery: the query serves the field's remaining segments degraded and
+ *    `quarantinedSegments()` reports the torn segment (loud once, counted
+ *    always, healable);
+ *  - a manifest-listed segment with NO bytes (gone on disk) quarantines the
+ *    same way.
  * Only genuine absence stays benign: querying a field that has no manifest at all
  * returns empty (nothing was ever written for it) — that is not a fault.
  */
 import { describe, it, expect, beforeEach } from 'vitest'
-import {
-  ColumnStore,
-  ColumnSegmentLoadError
-} from '../../../../src/indexes/columnStore/ColumnStore.js'
+import { ColumnStore } from '../../../../src/indexes/columnStore/ColumnStore.js'
 import { MemoryStorage } from '../../../../src/storage/adapters/memoryStorage.js'
 import { EntityIdMapper } from '../../../../src/utils/entityIdMapper.js'
 
@@ -80,30 +83,44 @@ describe('ColumnStore segment-load faults surface loudly, absence stays benign (
     return s
   }
 
-  it('propagates a storage IO fault verbatim — not [] and not a ColumnSegmentLoadError', async () => {
+  it('propagates a storage IO fault verbatim — not [] and not a quarantine (a present-but-unreadable segment is not torn)', async () => {
     storage.faultMode = 'io'
     const store = await reopen()
     await expect(store.filter('createdAt', 300)).rejects.toMatchObject({
       code: 'EIO'
     })
+    // An IO fault is NOT quarantined — the segment may be fine once the disk
+    // recovers; only torn/absent bytes enter the ledger.
+    expect(store.quarantinedSegments('createdAt')).toEqual([])
     await store.close()
   })
 
-  it('throws ColumnSegmentLoadError when a manifest-listed segment is undecodable', async () => {
+  it('QUARANTINES an undecodable manifest-listed segment at discovery — the query serves degraded, the ledger names the tear', async () => {
     storage.faultMode = 'corrupt'
     const store = await reopen()
-    await expect(
-      store.sortTopK('createdAt', 'desc', 10)
-    ).rejects.toBeInstanceOf(ColumnSegmentLoadError)
+    // Degraded-announced serve: the field's only segment is torn, so the
+    // result is empty — but the query completes instead of throwing.
+    const sorted = await store.sortTopK('createdAt', 'desc', 10)
+    expect(sorted).toEqual([])
+    const ledger = store.quarantinedSegments('createdAt')
+    expect(ledger).toHaveLength(1)
+    expect(ledger[0].error).toMatch(/decode failed/)
+    expect(ledger[0].hits).toBeGreaterThanOrEqual(1)
+    // Subsequent queries keep serving (skip + count), never a throw.
+    const hitsBefore = ledger[0].hits
+    await expect(store.filter('createdAt', 300)).resolves.toBeDefined()
+    expect(store.quarantinedSegments('createdAt')[0].hits).toBeGreaterThan(hitsBefore)
     await store.close()
   })
 
-  it('throws ColumnSegmentLoadError when a manifest-listed segment has no loadable bytes', async () => {
+  it('QUARANTINES a manifest-listed segment with no loadable bytes — degraded serve, ledger entry, never a throw', async () => {
     storage.faultMode = 'missing'
     const store = await reopen()
-    await expect(
-      store.rangeQuery('createdAt', 100, 500)
-    ).rejects.toBeInstanceOf(ColumnSegmentLoadError)
+    const bitmap = await store.rangeQuery('createdAt', 100, 500)
+    expect(bitmap.size).toBe(0)
+    const ledger = store.quarantinedSegments('createdAt')
+    expect(ledger).toHaveLength(1)
+    expect(ledger[0].error).toMatch(/no loadable bytes/)
     await store.close()
   })