docs(research): sync wiki with absorption work; rotate stale log entries

- log.md: correct the 2026-06-17 entry — DELAR-1/2/3 implemented (was
  stale "planned, not authorised"); SENARE still gated.
- Rotate the three 2026-06-15 Issue #32 tooling feats (gallery / ledger /
  overlap detector) to log-archive part 1 to stay under the 28KB log bound.
- handoff.md: add a Recent Changes line for the NU+DELAR absorption;
  compress two already-archived 06-15 lines to hold the 300-line bound.

Docs-only; no code/facit/signal/auto-fib touch. Repo-bounds green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: JohnCCarter

docs/research_wiki/handoff.md

@@ -22,6 +22,12 @@ append-only trail lives in [log.md](log.md).
 
 ## Recent Changes
 
+- **2026-06-17 External-pattern-scan absorption — NU+DELAR done** (branch `feature/research-fib`;
+  plan `clever-yawning-catmull.md`). NU (docs): standing
+  [prereg addendum](reviews/horizontal-structure-prereg-addendum-20260617.md) (random-walk control /
+  anytime-valid re-looks / embargo named as purged-CV). DELAR (code): synthetic RW baseline,
+  uncertainty-ordered worklist (`--by-uncertainty`), fail-closed swing-label validation. SENARE gated.
+  No facit/signal/auto-fib touch.
 - **2026-06-16 BTC/Fib studies — BOTH NULL, reviewed PASS / CLOSED** (commit `f4e96f1`, Lean Fib).
   **(1) Behaviour event study:** fib ≈ placebo ≈ swing (4h reject 0.78/0.80/0.84, p=0.63/0.19).
   **(2) Context-conditioned** (continuous MFE−MAE, rank-perm + Holm; contexts trend + deep
@@ -55,13 +61,10 @@ append-only trail lives in [log.md](log.md).
   2017-12-28T20:00 @ 13611 → 2017-12-28T08:00 @ 13145 (candidate_1). Only anchor_a + levels
   changed; anchor_b/fib_id unchanged; guard PASS; ledger candidate → corrected. Closes the
   declutter→correction→ledger track. [Report](reviews/btc-4h-fib-20171228-correction-20260615.md).
-- **2026-06-15 Issue #32 top-ROI tooling — DONE** (all stdlib, report-only): single-fib declutter
-  edit-mode (`labeling/tool.py --edit-fib-id`), overlap/dedup detector (`research/overlap_detector.py`,
-  22 candidates), source-quality review ledger (`research/review_ledger.py`), static HTML artifact
-  gallery (`research/artifact_gallery.py`). Detail in [log.md](log.md) + reviews.
-- **2026-06-15 4H Tier 1 + Tier 2 visual reviews — complete** (entries archived to
-  [log part 1](log-archive-btc-postreset-part1.md)): 11 groups mapped, Tier 2 sample-pass; one
-  correction-candidate `20171228` (now corrected). Corpus declared clean below.
+- **2026-06-15 Issue #32 top-ROI tooling — DONE** (declutter edit-mode, overlap detector, review
+  ledger, artifact gallery; all stdlib). Detail in [log part 1](log-archive-btc-postreset-part1.md).
+- **2026-06-15 4H Tier 1 + Tier 2 visual reviews — complete** (11 groups; `20171228` corrected;
+  corpus clean). Archived to [log part 1](log-archive-btc-postreset-part1.md).
 - **2026-06-08→06-12 source-fib milestones (archived)** — 1M/1W/1D/4H source phases, reaction
   reviews (1816 interactions), 4H Tier 1 maps, Addendum 2 golden-zone retirement (#30), log-scale
   + profile fix + monthly-first reset. Detail: [log post-reset part 1](log-archive-btc-postreset-part1.md)

docs/research_wiki/log-archive-btc-postreset-part1.md

@@ -133,3 +133,48 @@ fib count. A 55-fib group (2021) can be mostly map-OK; a 103-fib group (2017_h2)
 maps; Tier 2 per-fib zoom windows will be narrower and more readable.
 
 **Next:** implement Tier 2 `fourh_source_fib_zoom.py`.
+
+## [2026-06-15] feat | Overlap/dedup detector + anchor convention (Issue #32 top-ROI #3)
+
+Added `research/overlap_detector.py` — stdlib-only, report-only detector. Each fib is a
+box in (time, log-price) space; per pair it computes time/price/box IoU + shared-anchor.
+Flags candidates (box_iou≥0.5 or shared anchor) for human review — never edits labels,
+never says "wrong". Fail-closed timeframe guard. Real run on 366 4H fibs: **22 candidate
+pairs, all sharing anchor_b** (no pure-geometric overlap ≥0.5) → dominant signal is
+sub-legs ending on the same swing, not duplicates. 20210110 pair confirmed (box_iou 0.51);
+2017_h2 cluster present; strongest near-dup 20250506 pair (box_iou 0.70); `20171228`
+correctly absent (anchor-quality issue, not duplication). Report:
+[`reviews/btc-4h-overlap-candidates-20260615.md`](reviews/btc-4h-overlap-candidates-20260615.md)
++ CSV. Anchor convention (body/close vs wick, observed not absolute) documented in
+[`labeling/HUMAN_FIB_ANNOTATION.md`](../labeling/HUMAN_FIB_ANNOTATION.md). 9 tests; ruff +
+full suite green (361 passed, 75.07% cov). No source labels changed; no new deps.
+
+## [2026-06-15] feat | Source-quality review ledger (Issue #32 top-ROI #2)
+
+Added `research/review_ledger.py` — stdlib-only helper (csv/hashlib/json; no new deps)
+that makes source-fib review verdicts machine-trackable. Flat CSV, controlled vocab
+(verdict ∈ ok/ok-with-note/watchlist/suspicious; status ∈ accepted/noted/open/
+correction-candidate/deferred/corrected), with a deterministic `source_hash`
+(`sha256:<16 hex>` of the fib JSON bytes) tying each verdict to the exact facit version.
+Generated the first ledger for the 4H Tier 2 sample-pass (8 rows) at
+[`reviews/ledgers/btc-4h-source-quality-ledger.csv`](reviews/ledgers/btc-4h-source-quality-ledger.csv);
+`fib_BTC-USD_4h_20171228T200000` represented as suspicious / correction-candidate. Schema
+doc: [`reviews/ledgers/README.md`](reviews/ledgers/README.md). 12 tests
+(`tests/research/test_review_ledger.py`: hash determinism, vocab validation,
+correction-candidate representable, roundtrip, header check); ruff + full suite green
+(352 passed, 74.91% cov). No source labels changed; CSV is committed text under `docs/`.
+
+## [2026-06-15] feat | Static HTML artifact gallery (Issue #32 top-ROI #1)
+
+Added `research/artifact_gallery.py` — stdlib-only helper (no new deps) that scans a
+review PNG directory and writes a self-contained `index.html` beside it (relative links,
+inline CSS/JS-free, clean+levels paired per item). Auto-detects both layouts: flat **map**
+output (`..._<label>_4h_<kind>.png`) and nested **zoom** output (`<scope>/<fib_id>/
+4h_<kind>.png`). Standalone helper by design — does **not** touch the render modules or
+the existing markdown `_write_index`. Output lands under `experiments/review/**`
+(gitignored; HTML not committed). 9 tests in `tests/research/test_artifact_gallery.py`
+(both layouts, relative-links-only, empty/missing dir, no-external-deps, markdown index
+untouched); ruff + full suite green (340 passed, 74.77% cov). Real smoke run: zoom (140
+items) + map galleries written; `git status` confirms `index.html` is ignored.
+
+Build: `python -m fibengine.research.artifact_gallery --root experiments/review/fourh_source_fib_zoom`.

docs/research_wiki/log.md

@@ -21,7 +21,10 @@ Types: `ingest`, `decision`, `review`, `question`, `maintenance`.
 NU block of the external pattern scan, docs-only:
 [horizontal-structure-prereg-addendum-20260617.md](reviews/horizontal-structure-prereg-addendum-20260617.md)
 (NU-1 random-walk control; NU-2 anytime-valid/e-Holm re-looks; NU-3 name embargo as purged-CV).
-DELAR/SENARE code planned, not authorised (`clever-yawning-catmull.md`).
+DELAR-1/2/3 since implemented (commits `ca5ae73`/`0b380e6`/`7b03837`): synthetic random-walk
+baseline (`research/synthetic_baseline.py`), uncertainty-ordered labeling worklist
+(`labeling/worklist.py --by-uncertainty`), fail-closed swing-label JSON validation
+(`validation/schemas.py`). SENARE still gated (`clever-yawning-catmull.md`).
 
 ## [2026-06-16] decision | BTC/Fib behaviour/backtest line — PAUSED / CLOSED (reviewed PASS)
 
@@ -355,53 +358,8 @@ ledger; anchor_a only, body/close convention) → (3) update ledger row candidat
 **Also evaluate-later:** chart-regression strategy (structural/hash vs pytest-mpl,
 binary-baseline / anti-blob question). 1H source labeling remains deferred.
 
-## [2026-06-15] feat | Overlap/dedup detector + anchor convention (Issue #32 top-ROI #3)
-
-Added `research/overlap_detector.py` — stdlib-only, report-only detector. Each fib is a
-box in (time, log-price) space; per pair it computes time/price/box IoU + shared-anchor.
-Flags candidates (box_iou≥0.5 or shared anchor) for human review — never edits labels,
-never says "wrong". Fail-closed timeframe guard. Real run on 366 4H fibs: **22 candidate
-pairs, all sharing anchor_b** (no pure-geometric overlap ≥0.5) → dominant signal is
-sub-legs ending on the same swing, not duplicates. 20210110 pair confirmed (box_iou 0.51);
-2017_h2 cluster present; strongest near-dup 20250506 pair (box_iou 0.70); `20171228`
-correctly absent (anchor-quality issue, not duplication). Report:
-[`reviews/btc-4h-overlap-candidates-20260615.md`](reviews/btc-4h-overlap-candidates-20260615.md)
-+ CSV. Anchor convention (body/close vs wick, observed not absolute) documented in
-[`labeling/HUMAN_FIB_ANNOTATION.md`](../labeling/HUMAN_FIB_ANNOTATION.md). 9 tests; ruff +
-full suite green (361 passed, 75.07% cov). No source labels changed; no new deps.
-
-## [2026-06-15] feat | Source-quality review ledger (Issue #32 top-ROI #2)
-
-Added `research/review_ledger.py` — stdlib-only helper (csv/hashlib/json; no new deps)
-that makes source-fib review verdicts machine-trackable. Flat CSV, controlled vocab
-(verdict ∈ ok/ok-with-note/watchlist/suspicious; status ∈ accepted/noted/open/
-correction-candidate/deferred/corrected), with a deterministic `source_hash`
-(`sha256:<16 hex>` of the fib JSON bytes) tying each verdict to the exact facit version.
-Generated the first ledger for the 4H Tier 2 sample-pass (8 rows) at
-[`reviews/ledgers/btc-4h-source-quality-ledger.csv`](reviews/ledgers/btc-4h-source-quality-ledger.csv);
-`fib_BTC-USD_4h_20171228T200000` represented as suspicious / correction-candidate. Schema
-doc: [`reviews/ledgers/README.md`](reviews/ledgers/README.md). 12 tests
-(`tests/research/test_review_ledger.py`: hash determinism, vocab validation,
-correction-candidate representable, roundtrip, header check); ruff + full suite green
-(352 passed, 74.91% cov). No source labels changed; CSV is committed text under `docs/`.
-
-## [2026-06-15] feat | Static HTML artifact gallery (Issue #32 top-ROI #1)
-
-Added `research/artifact_gallery.py` — stdlib-only helper (no new deps) that scans a
-review PNG directory and writes a self-contained `index.html` beside it (relative links,
-inline CSS/JS-free, clean+levels paired per item). Auto-detects both layouts: flat **map**
-output (`..._<label>_4h_<kind>.png`) and nested **zoom** output (`<scope>/<fib_id>/
-4h_<kind>.png`). Standalone helper by design — does **not** touch the render modules or
-the existing markdown `_write_index`. Output lands under `experiments/review/**`
-(gitignored; HTML not committed). 9 tests in `tests/research/test_artifact_gallery.py`
-(both layouts, relative-links-only, empty/missing dir, no-external-deps, markdown index
-untouched); ruff + full suite green (340 passed, 74.77% cov). Real smoke run: zoom (140
-items) + map galleries written; `git status` confirms `index.html` is ignored.
-
-Build: `python -m fibengine.research.artifact_gallery --root experiments/review/fourh_source_fib_zoom`.
-
-
 > **2026-06-11→06-12 entries** (1M reaction-review, 1W/1D/4H source phases, 4H Tier 1
-> design/maps) **and the 2026-06-15 4H Tier 1/Tier 2 visual-review entries** archived to
+> design/maps), **the 2026-06-15 4H Tier 1/Tier 2 visual-review entries**, **and the
+> 2026-06-15 Issue #32 tooling feats** (gallery / ledger / overlap detector) archived to
 > [post-reset part 1](log-archive-btc-postreset-part1.md).