SniperGold_ML/docs/P3_GOV_2_REPOSITORY_TRUTH_AUDIT.md

# P3-GOV-2 — REPOSITORY TRUTH AUDIT

```text
Document type : GOVERNANCE / REPOSITORY TRUTH AUDIT
Session       : P3-GOV-2 (REPOSITORY TRUTH AUDIT & AI ASSISTANT CLAIM RECONCILIATION)
Date          : 2026-08-28
Read-only     : this audit did NOT modify any research artifact, checkpoint,
                MQL5, ML output, Tickstory/dukascopy data, or historical report.
Final class   : B — MOSTLY RECONCILED WITH DOCUMENTED GAPS
Principal     : REPOSITORY TRUTH > AI ASSISTANT CLAIMS
```

## A. Repository identity

```text
Authoritative workspace:
  D:\TradingTerminal\HFM Metatrader 5\MQL5\Shared Projects\SniperGold_ML_authoritative

Remote:
  https://forge.mql5.io/chiki2bum2/SniperGold_ML.git
```

## B. Actual Git state

```text
local HEAD  : 7e7b6726cf04de85148c3005ed00dc5460da901e
remote HEAD : 7e7b6726cf04de85148c3005ed00dc5460da901e
branch      : main
HEAD message: "test: isolate and repair P3-S25.1D interruption recovery verifier"
working tree: tracked-files CLEAN (only the new P3-GOV-2 governance-audit namespace is
              untracked; committed at close-out)
stash       : NONE
tags        : (none)
origin      : https://forge.mql5.io/chiki2bum2/SniperGold_ML.git   (expected)
```

**Baseline: PASS** (local HEAD == origin/main; branch main; no stash; expected origin).

## C. Phase reconciliation

Every claimed phase Close-SHA was resolved to a full commit object and verified for
existence, parent chain, and reachability from local HEAD and origin/main. Values below
are the authoritative resolved SHAs (machine-recorded in
`ml/p3/governance/p3_gov2_repository_truth_audit/output/claim_reconciliation.json`).

| Phase        | Claimed state            | Resolved commit (authoritative)                 | Local | origin/main | Rpt | H/O | NS present |
| ------------ | ------------------------ | ----------------------------------------------- | ----- | ---------- | --- | --- | ---------- |
| P3-S18       | INCONCLUSIVE             | `a1283929d8b92fdac16883143f509907ee424dc2`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S19       | VERIFIED                 | `84b0e56fc1e6ba2be503f462c4ef1a58e0830915`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S20       | FROZEN weak signal       | `7c9602ca4964f42f4ae8b7ac1ea14ab5e49e9176`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S21.1     | descriptive coefs        | `49f2f03f85a7a39ba0da115d98e103a98595509c`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S21.2     | B                        | `da54c9b399eab92f696cbc38d4b4e5d52a764c13`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S21.R     | NO SILENT BUG            | `ab7d444aa475d2a232d30d676820a7bdc407112f`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S22       | B passed w/ limits       | `6e553520e019e492da03831da07f850e4e377dbe`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S22.1     | ADJ-1/ADJ-2              | `ef80fc31b44e6bcf90d4999bffe80b84f0df87da`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S22.2     | ADJ-1/ADJ-2              | `a033e2705fce8b0240a3800050db14b2106c5068`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S22.3     | C - material change      | `0ee811e011c149b15d0586517bf24aa616c9e544`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S22.4     | B VALIDATED              | `e633baa96c75a2e8830628ce92a627aee4120cce`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S23       | B weak/inconsistent      | `a6a804b7ced141f099214dec5b4f705d94835e35`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S24       | D - edge NOT shown       | `8f1c80abe72a8db35a5fd1582c68e73033d0843a`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S25       | E-UNRESOLVED             | `84465a2e2462801bd906a98a4047ea4ea7568634`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S25.1     | STARTED                  | `8d631ed58d1bb86d02513b295a72a9c352ddb4be`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S25.1.R   | COMPLETE                 | `390bd9e85c5d32d50848dc3713fd4304d9e6779b`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S25.1 SB  | A (chunks 0-10)          | `6241383fad6d3405a45472b7e8c6370ee70ae025`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S25.1C    | A - chunk size verified  | `0b38e96bf8d39afdb6b5ce2069d1b08b5993b619`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-S25.1D    | COMPLETE (multi-chunk)   | `7e7b6726cf04de85148c3005ed00dc5460da901e`      | ✓     | ✓          | ✓   | folded-into-DR | ✓ |
| P3-S25.1D-R  | COMPLETE (A)             | `7e7b6726cf04de85148c3005ed00dc5460da901e`      | ✓     | ✓          | ✓   | ✓   | ✓ |
| P3-GOV-1     | CLOSED (B)               | `49b174eb4103dc5707c8411c1b3b6c8cec3579be`      | ✓     | ✓          | ✓   | ✓   | ✓ |

**Material commit conflicts: 0.** Every claimed SHA exists as a commit, is linked in the
parent chain, and is reachable from `origin/main`.

Interpretation note (two-commit convention): for phases whose Close-SHA is the phase's
documentation tip (S18, S19, S20, S21.x, S22, S22.1, S22.2 — the doc recording the phase
result), the substantive research commit is its parent. For machine-evidence phases
(S22.3, S22.4, S23, S24, S25, S25.1, S25.1C, S25.1D/D-R, GOV-1) the Close-SHA is the
substantive/evidence commit (research or combined documentation+research commit). No
claimed SHA was found to be a bare doc-tip mislabeled as substantive, and no substantive
commit was mislabeled as the repo tip.

## D. Documentation / artifact contradictions

| # | Finding | Severity |
|---| ------- | -------- |
| D1 | `CURRENT_DECISION_GATE.md` still authorizes "CHUNK 9 ONLY / offset 113,246,416" whereas the real checkpoint is chunk 10 committed / next 11 / offset 163,578,081 | minor-doc |
| D2 | `README.md` CURRENT-status block pinned at "P3-S25.1C COMPLETE / conclusion D" while `CURRENT_PROJECT_STATE.md` has advanced to "P3-S25.1 Stage-B CHUNK 10 COMPLETE" | minor-doc |
| D3 | `CURRENT_PROJECT_STATE.md` states `run chunk_size = 25165824 (24 MiB)` but the authoritative checkpoint top-level `chunk_size` field is `12582912` (12 MiB, the initial-run value); per-chunk manifests show 24 MiB for chunks 9-10 only | minor-doc (canonical field record) |
| D4 | `PROJECT_PHASE_INDEX.md` header "CURRENT AUTHORITATIVE phase = P3-S25.1C COMPLETE (HEAD 0b38e96b...)" is stale after chunks 9-10 and S25.1D | minor-doc |
| D5 | GOV-1 handover records final SHA `983adbd9...`; GOV-1 tip commit is actually `49b174e` (the commit that wrote that record). `983adbd` is the last substantive GOV-1 classification commit. | minor-doc |

None of D1-D5 contradict research/commit evidence; all are documentation/narrative staleness.

## E. P3-S25.1 current real checkpoint truth

```text
run_id                    P3_S251_REAL_TICK_RUN_001
source_path               D:\...\MQL5\Files\XAUUSD_mt5_ticks.csv
source_sha256             5252ce8f5b0e7d286b71a3172c7ef9c1009c8e4b1254ac67400a803fcd85cba1
source_size               34,473,661,010 bytes (34.5 GB)
status                    CHUNK_COMMITTED
last_completed_chunk      10
current_chunk             10
next chunk                11
next byte offset          163,578,081
cumulative rows           3,661,058
cumulative M15            18,579
cumulative M30            9,966
last_timestamp            1,078,220,689
checkpoint file           ml/p3/p3_s251_external_ingest/output/s251_real_checkpoint.json
checkpoint sha256         47da8f020426bec67e9f7865519e900a28c286ba1d13a15bee6a8c0f5307afc7
full ingestion            NOT COMPLETE (34.5 GB / 722,941,870 rows; ~163.6 MB ingested)
```

Chunk-10 manifest reconciliation: rows 563,266 -> cumulative 3,661,058; M15 2,462 -> 18,579;
M30 1,317 -> 9,966; byte end 163,578,081. Exact. `CHUNK 11 NOT PROCESSED`.

## F. P3-S25.1C truth

```text
SAFE_MAX_CHUNK_SIZE = 24 MiB (25165824 bytes)
chunk-size invariance = PASS (12/24/48/64 MiB byte-identical; all diffs_vs_reference empty)
boundary audit        = PASS (boundary_audit_ok = true; 5 scenarios all equal)
mutation detection    = 10/10 (target_ratio 10/10, detected_ratio 10/10, all_detected true)
48/64 MiB tested      = yes; NOT promoted to SAFE_MAX; SAFE_MAX stayed 24 MiB
                        (evidence-based safe cap; invariance proven up to 64 MiB)
```

## G. P3-S25.1D-R truth

```text
recovery manifest status:
  original_failure_isolated           true
  corrected_resume_matches_continuous true
  reproducible_run1_equals_run2       true
  mutation_detection                  8/8
  real_checkpoint_unchanged           true
  chunk11_processed                   false
decision gate (isolated path)        A - VERIFIER DEFECT ISOLATED AND FIX VERIFIED
boundary defect     RESUME_BOUNDARY_MISMATCH; earliest divergence chunk 1
                    (continuous actual byte_end 25,165,841 vs obsolete nominal 25,165,824)
root cause          DISPOSABLE verifier driver rebuilt resume from obsolete nominal
                    split_atomic 24 MiB boundaries instead of actual line-safe consumed
                    byte_end. NOT the production parser/aggregator/checkpoint.
previous failing output: OVERWRITTEN (the interrupted session overwrote the original
                    failing JSON before stopping; recorded, not recoverable on disk)
SAFE_CHUNKS_PER_TURN: NOT SET (broader P3-S25.1D multi-chunk benchmark remains OPEN)
```

## H. Documentation narrative conflicts

See D1-D5. All documentation-only; no historical report / machine evidence / commit conflict.

## I. Untracked / interrupted work

```text
The only untracked path is ml/p3/governance/ (the P3-GOV-2 audit namespace).
No CHUNK-11 or interruption-temp remnants present (previous-session untracked S25.1D
artifacts were committed by HEAD 7e7b672).
checkpoint before == after : 47da8f02... == 47da8f02...   (TRUE)
```

## J. Trust grading (repository provenance, T0-T5)

| Phase(s) | Grade | Basis |
| -------- | ----- | ----- |
| P3-S18..S20, S21.x, S22, S22.1/22.2 | T3 | committed, remote-reachable, report+handover present |
| P3-S22.3, P3-S22.4, P3-S23, P3-S24, P3-S25 | T4 | above + machine JSON values verified against claims |
| P3-S25.1, P3-S25.1.R, P3-S25.1 Stage-B | T4 | above + real checkpoint verified (rows/M15/M30/offset) |
| P3-S25.1C | T5 | report + JSON + handover + commit reconciled (invariance/10-10/boundary) |
| P3-S25.1D, P3-S25.1D-R | T5 | report + recovery JSON + handover + commit reconciled (8/8, checkpoint unchanged) |
| P3-GOV-1 | T4 | governance doc/self-audit JSON present, handover present; final-SHA nuance D5 |

## K. Required remediation (recommendations ONLY; nothing changed this session)

1. Update `CURRENT_DECISION_GATE.md` next-action to `CHUNK 11 ONLY @ offset 163,578,081` (or the
   owner decision from S25.1D-R handover: close-out of S25.1D benchmark vs continue ingestion).
2. Align `README.md` CURRENT-status block to `CURRENT_PROJECT_STATE.md` (CHUNK 10 COMPLETE).
3. Update `PROJECT_PHASE_INDEX.md` CURRENT header to actual tip and add S25.1D/D-R/GOV-1 rows.
4. Record in the S25.1 checkpoint/state docs that the top-level `chunk_size` field is the
   initial 12 MiB run value; effective per-chunk size lives in each chunk manifest.
5. GOV-1 handover: state that GOV-1 tip = `49b174e` (substantive classification commit `983adbd`).

## L. Final classification

```text
B — MOSTLY RECONCILED WITH DOCUMENTED GAPS

Core claims (phase chain, Git SHAs, handover chain, machine evidence, final-state values)
are supported by and consistent with the authoritative repository and origin/main.
Remaining gaps are non-critical documentation/narrative staleness (README, DECISION_GATE,
PROJECT_PHASE_INDEX) and a GOV-1 final-SHA recording nuance. No research claim is
contradicted by repository or artifact evidence.
```

## Audit integrity self-check

```text
checkpoint immutability : TRUE  (47da8f02... before == after)
reproducibility         : TRUE  (two semantic runs byte-identical)
auditor mutation test   : 10/10 disposable false claims DETECTED
machine outputs         : ml/p3/governance/p3_gov2_repository_truth_audit/output/*.json
```

*End of P3-GOV-2 repository truth audit.*
audit: reconcile AI Assistant claims with repository truth 2026-08-28 21:15:19 +07:00			`# P3-GOV-2 — REPOSITORY TRUTH AUDIT`

			```text
			`Document type : GOVERNANCE / REPOSITORY TRUTH AUDIT`
			`Session : P3-GOV-2 (REPOSITORY TRUTH AUDIT & AI ASSISTANT CLAIM RECONCILIATION)`
			`Date : 2026-08-28`
			`Read-only : this audit did NOT modify any research artifact, checkpoint,`
			`MQL5, ML output, Tickstory/dukascopy data, or historical report.`
			`Final class : B — MOSTLY RECONCILED WITH DOCUMENTED GAPS`
			`Principal : REPOSITORY TRUTH > AI ASSISTANT CLAIMS`
			```

			`## A. Repository identity`

			```text
			`Authoritative workspace:`
			`D:\TradingTerminal\HFM Metatrader 5\MQL5\Shared Projects\SniperGold_ML_authoritative`

			`Remote:`
			`https://forge.mql5.io/chiki2bum2/SniperGold_ML.git`
			```

			`## B. Actual Git state`

			```text
			`local HEAD : 7e7b6726cf04de85148c3005ed00dc5460da901e`
			`remote HEAD : 7e7b6726cf04de85148c3005ed00dc5460da901e`
			`branch : main`
			`HEAD message: "test: isolate and repair P3-S25.1D interruption recovery verifier"`
			`working tree: tracked-files CLEAN (only the new P3-GOV-2 governance-audit namespace is`
			`untracked; committed at close-out)`
			`stash : NONE`
			`tags : (none)`
			`origin : https://forge.mql5.io/chiki2bum2/SniperGold_ML.git (expected)`
			```

			`Baseline: PASS (local HEAD == origin/main; branch main; no stash; expected origin).`

			`## C. Phase reconciliation`

			`Every claimed phase Close-SHA was resolved to a full commit object and verified for`
			`existence, parent chain, and reachability from local HEAD and origin/main. Values below`
			`are the authoritative resolved SHAs (machine-recorded in`
			`ml/p3/governance/p3_gov2_repository_truth_audit/output/claim_reconciliation.json`).

			`\| Phase \| Claimed state \| Resolved commit (authoritative) \| Local \| origin/main \| Rpt \| H/O \| NS present \|`
			`\| ------------ \| ------------------------ \| ----------------------------------------------- \| ----- \| ---------- \| --- \| --- \| ---------- \|`
			\| P3-S18 \| INCONCLUSIVE \| `a1283929d8b92fdac16883143f509907ee424dc2` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S19 \| VERIFIED \| `84b0e56fc1e6ba2be503f462c4ef1a58e0830915` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S20 \| FROZEN weak signal \| `7c9602ca4964f42f4ae8b7ac1ea14ab5e49e9176` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S21.1 \| descriptive coefs \| `49f2f03f85a7a39ba0da115d98e103a98595509c` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S21.2 \| B \| `da54c9b399eab92f696cbc38d4b4e5d52a764c13` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S21.R \| NO SILENT BUG \| `ab7d444aa475d2a232d30d676820a7bdc407112f` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S22 \| B passed w/ limits \| `6e553520e019e492da03831da07f850e4e377dbe` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S22.1 \| ADJ-1/ADJ-2 \| `ef80fc31b44e6bcf90d4999bffe80b84f0df87da` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S22.2 \| ADJ-1/ADJ-2 \| `a033e2705fce8b0240a3800050db14b2106c5068` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S22.3 \| C - material change \| `0ee811e011c149b15d0586517bf24aa616c9e544` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S22.4 \| B VALIDATED \| `e633baa96c75a2e8830628ce92a627aee4120cce` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S23 \| B weak/inconsistent \| `a6a804b7ced141f099214dec5b4f705d94835e35` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S24 \| D - edge NOT shown \| `8f1c80abe72a8db35a5fd1582c68e73033d0843a` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S25 \| E-UNRESOLVED \| `84465a2e2462801bd906a98a4047ea4ea7568634` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S25.1 \| STARTED \| `8d631ed58d1bb86d02513b295a72a9c352ddb4be` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S25.1.R \| COMPLETE \| `390bd9e85c5d32d50848dc3713fd4304d9e6779b` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S25.1 SB \| A (chunks 0-10) \| `6241383fad6d3405a45472b7e8c6370ee70ae025` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S25.1C \| A - chunk size verified \| `0b38e96bf8d39afdb6b5ce2069d1b08b5993b619` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-S25.1D \| COMPLETE (multi-chunk) \| `7e7b6726cf04de85148c3005ed00dc5460da901e` \| ✓ \| ✓ \| ✓ \| folded-into-DR \| ✓ \|
			\| P3-S25.1D-R \| COMPLETE (A) \| `7e7b6726cf04de85148c3005ed00dc5460da901e` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|
			\| P3-GOV-1 \| CLOSED (B) \| `49b174eb4103dc5707c8411c1b3b6c8cec3579be` \| ✓ \| ✓ \| ✓ \| ✓ \| ✓ \|

			`Material commit conflicts: 0. Every claimed SHA exists as a commit, is linked in the`
			parent chain, and is reachable from `origin/main`.

			`Interpretation note (two-commit convention): for phases whose Close-SHA is the phase's`
			`documentation tip (S18, S19, S20, S21.x, S22, S22.1, S22.2 — the doc recording the phase`
			`result), the substantive research commit is its parent. For machine-evidence phases`
			`(S22.3, S22.4, S23, S24, S25, S25.1, S25.1C, S25.1D/D-R, GOV-1) the Close-SHA is the`
			`substantive/evidence commit (research or combined documentation+research commit). No`
			`claimed SHA was found to be a bare doc-tip mislabeled as substantive, and no substantive`
			`commit was mislabeled as the repo tip.`

			`## D. Documentation / artifact contradictions`

			`\| # \| Finding \| Severity \|`
			`\|---\| ------- \| -------- \|`
			\| D1 \| `CURRENT_DECISION_GATE.md` still authorizes "CHUNK 9 ONLY / offset 113,246,416" whereas the real checkpoint is chunk 10 committed / next 11 / offset 163,578,081 \| minor-doc \|
			\| D2 \| `README.md` CURRENT-status block pinned at "P3-S25.1C COMPLETE / conclusion D" while `CURRENT_PROJECT_STATE.md` has advanced to "P3-S25.1 Stage-B CHUNK 10 COMPLETE" \| minor-doc \|
			\| D3 \| `CURRENT_PROJECT_STATE.md` states `run chunk_size = 25165824 (24 MiB)` but the authoritative checkpoint top-level `chunk_size` field is `12582912` (12 MiB, the initial-run value); per-chunk manifests show 24 MiB for chunks 9-10 only \| minor-doc (canonical field record) \|
			\| D4 \| `PROJECT_PHASE_INDEX.md` header "CURRENT AUTHORITATIVE phase = P3-S25.1C COMPLETE (HEAD 0b38e96b...)" is stale after chunks 9-10 and S25.1D \| minor-doc \|
			\| D5 \| GOV-1 handover records final SHA `983adbd9...`; GOV-1 tip commit is actually `49b174e` (the commit that wrote that record). `983adbd` is the last substantive GOV-1 classification commit. \| minor-doc \|

			`None of D1-D5 contradict research/commit evidence; all are documentation/narrative staleness.`

			`## E. P3-S25.1 current real checkpoint truth`

			```text
			`run_id P3_S251_REAL_TICK_RUN_001`
			`source_path D:\...\MQL5\Files\XAUUSD_mt5_ticks.csv`
			`source_sha256 5252ce8f5b0e7d286b71a3172c7ef9c1009c8e4b1254ac67400a803fcd85cba1`
			`source_size 34,473,661,010 bytes (34.5 GB)`
			`status CHUNK_COMMITTED`
			`last_completed_chunk 10`
			`current_chunk 10`
			`next chunk 11`
			`next byte offset 163,578,081`
			`cumulative rows 3,661,058`
			`cumulative M15 18,579`
			`cumulative M30 9,966`
			`last_timestamp 1,078,220,689`
			`checkpoint file ml/p3/p3_s251_external_ingest/output/s251_real_checkpoint.json`
			`checkpoint sha256 47da8f020426bec67e9f7865519e900a28c286ba1d13a15bee6a8c0f5307afc7`
			`full ingestion NOT COMPLETE (34.5 GB / 722,941,870 rows; ~163.6 MB ingested)`
			```

			`Chunk-10 manifest reconciliation: rows 563,266 -> cumulative 3,661,058; M15 2,462 -> 18,579;`
			M30 1,317 -> 9,966; byte end 163,578,081. Exact. `CHUNK 11 NOT PROCESSED`.

			`## F. P3-S25.1C truth`

			```text
			`SAFE_MAX_CHUNK_SIZE = 24 MiB (25165824 bytes)`
			`chunk-size invariance = PASS (12/24/48/64 MiB byte-identical; all diffs_vs_reference empty)`
			`boundary audit = PASS (boundary_audit_ok = true; 5 scenarios all equal)`
			`mutation detection = 10/10 (target_ratio 10/10, detected_ratio 10/10, all_detected true)`
			`48/64 MiB tested = yes; NOT promoted to SAFE_MAX; SAFE_MAX stayed 24 MiB`
			`(evidence-based safe cap; invariance proven up to 64 MiB)`
			```

			`## G. P3-S25.1D-R truth`

			```text
			`recovery manifest status:`
			`original_failure_isolated true`
			`corrected_resume_matches_continuous true`
			`reproducible_run1_equals_run2 true`
			`mutation_detection 8/8`
			`real_checkpoint_unchanged true`
			`chunk11_processed false`
			`decision gate (isolated path) A - VERIFIER DEFECT ISOLATED AND FIX VERIFIED`
			`boundary defect RESUME_BOUNDARY_MISMATCH; earliest divergence chunk 1`
			`(continuous actual byte_end 25,165,841 vs obsolete nominal 25,165,824)`
			`root cause DISPOSABLE verifier driver rebuilt resume from obsolete nominal`
			`split_atomic 24 MiB boundaries instead of actual line-safe consumed`
			`byte_end. NOT the production parser/aggregator/checkpoint.`
			`previous failing output: OVERWRITTEN (the interrupted session overwrote the original`
			`failing JSON before stopping; recorded, not recoverable on disk)`
			`SAFE_CHUNKS_PER_TURN: NOT SET (broader P3-S25.1D multi-chunk benchmark remains OPEN)`
			```

			`## H. Documentation narrative conflicts`

			`See D1-D5. All documentation-only; no historical report / machine evidence / commit conflict.`

			`## I. Untracked / interrupted work`

			```text
			`The only untracked path is ml/p3/governance/ (the P3-GOV-2 audit namespace).`
			`No CHUNK-11 or interruption-temp remnants present (previous-session untracked S25.1D`
			`artifacts were committed by HEAD 7e7b672).`
			`checkpoint before == after : 47da8f02... == 47da8f02... (TRUE)`
			```

			`## J. Trust grading (repository provenance, T0-T5)`

			`\| Phase(s) \| Grade \| Basis \|`
			`\| -------- \| ----- \| ----- \|`
			`\| P3-S18..S20, S21.x, S22, S22.1/22.2 \| T3 \| committed, remote-reachable, report+handover present \|`
			`\| P3-S22.3, P3-S22.4, P3-S23, P3-S24, P3-S25 \| T4 \| above + machine JSON values verified against claims \|`
			`\| P3-S25.1, P3-S25.1.R, P3-S25.1 Stage-B \| T4 \| above + real checkpoint verified (rows/M15/M30/offset) \|`
			`\| P3-S25.1C \| T5 \| report + JSON + handover + commit reconciled (invariance/10-10/boundary) \|`
			`\| P3-S25.1D, P3-S25.1D-R \| T5 \| report + recovery JSON + handover + commit reconciled (8/8, checkpoint unchanged) \|`
			`\| P3-GOV-1 \| T4 \| governance doc/self-audit JSON present, handover present; final-SHA nuance D5 \|`

			`## K. Required remediation (recommendations ONLY; nothing changed this session)`

			1. Update `CURRENT_DECISION_GATE.md` next-action to `CHUNK 11 ONLY @ offset 163,578,081` (or the
			`owner decision from S25.1D-R handover: close-out of S25.1D benchmark vs continue ingestion).`
			2. Align `README.md` CURRENT-status block to `CURRENT_PROJECT_STATE.md` (CHUNK 10 COMPLETE).
			3. Update `PROJECT_PHASE_INDEX.md` CURRENT header to actual tip and add S25.1D/D-R/GOV-1 rows.
			4. Record in the S25.1 checkpoint/state docs that the top-level `chunk_size` field is the
			`initial 12 MiB run value; effective per-chunk size lives in each chunk manifest.`
			5. GOV-1 handover: state that GOV-1 tip = `49b174e` (substantive classification commit `983adbd`).

			`## L. Final classification`

			```text
			`B — MOSTLY RECONCILED WITH DOCUMENTED GAPS`

			`Core claims (phase chain, Git SHAs, handover chain, machine evidence, final-state values)`
			`are supported by and consistent with the authoritative repository and origin/main.`
			`Remaining gaps are non-critical documentation/narrative staleness (README, DECISION_GATE,`
			`PROJECT_PHASE_INDEX) and a GOV-1 final-SHA recording nuance. No research claim is`
			`contradicted by repository or artifact evidence.`
			```

			`## Audit integrity self-check`

			```text
			`checkpoint immutability : TRUE (47da8f02... before == after)`
			`reproducibility : TRUE (two semantic runs byte-identical)`
			`auditor mutation test : 10/10 disposable false claims DETECTED`
			`machine outputs : ml/p3/governance/p3_gov2_repository_truth_audit/output/*.json`
			```

			`End of P3-GOV-2 repository truth audit.`