Resonac Holdings (4004.T) — Investment Tree v1

Stage 7 essay (Phase 2B). R2-VERIFIED 2026-05-30 against primary-source FY2025 + Q1 FY2026 決算短信 (kessan tanshin) from Resonac IR. Date: 2026-05-03 · R2 re-anchor 2026-05-30 · Anchor price: SHWDY $113.68 (OTC ADR, 1:1 ≈ ordinary; 4004.T ¥18,720, both verified live 2026-05-30) · Market cap ~$21B / ~¥3.1-3.3T · Trailing P/E ~95x (impairment-depressed) / forward ~40x on guided FY2026 EPS ¥425.45 · Currency: JPY/USD Archetype: AI-back-end-semiconductor-materials leader — CONFIRMED, but fully valued post a ~5-6× re-rating

⚠️ R2 RE-ANCHOR (2026-05-30) — the most dramatic mis-anchor in the entire sweep, and a thesis REFRAME. The 2026-05-03 scaffold was an admitted "Tier-C directional sketch" anchored at ¥4,250 and framed Resonac as a speculative ABF second-source candidate. Primary sources reveal two things the sketch got wrong: - Price: scaffold ¥4,250 → real ¥18,720 TSE / $113.68 ADR — a +340% under-anchor. The 52-week range is ¥3,037–¥20,495: the stock ~5-6×'d as the AI-materials thesis confirmed. The scaffold anchored near the bottom of the run. (Prices verified live 2026-05-30 via Yahoo; SHWDY is an UNSPONSORED, thin OTC ADR — liquidity is a real sizing constraint.) - Thesis reframe: the engine is back-end AI-PACKAGING materials, not the speculative ABF-substitute story. FY2025 Semiconductor & Electronic Materials segment revenue ¥506.3B (+13.7%), record core operating profit ¥108.4B (+47%) — ~37.6% of revenue but >99% of total core operating profit. AI-share of back-end materials rose ~10%→~20% YoY. Q1 FY2026 accelerating: segment revenue +21.1%, profit +73.7%, back-end at a record quarterly high; consolidated core OP +126% YoY, H1 guidance raised 40%. The AI-materials thesis the scaffold called "speculative" has CONFIRMED and re-rated. - The catch — it's no longer cheap, and it's a single-engine, leveraged story. At ~40× forward (trailing ~95× is impairment-distorted) the easy money is gone. FY2025 GAAP net income ¥29.0B (−60.5%) on divestiture impairments; net debt ~¥681-707B (post Showa Denko Materials/Hitachi Chemical acquisition, deleveraging); the rest of the portfolio (Chemicals, Crasus, Mobility) is breakeven-to-loss. Net: h0 55→68%, xii 60→73%, archetype de-fragilized (speculative-fragile → cyclical-macro-sensitive). But the verdict stays cautious on VALUATION grounds: Watch / Hold-small — don't chase a 40× forward multiple after a 6× run. See evidence_2026-05-30.jsonl + update_2026-05-30.md. This R2 promotes RESONAC from a Tier-C "AJNMY-threat sketch" to a primary-source standalone memo (exactly the rebuild update_2026-05-25.md called for).

0. Company Fundamentals — what Resonac is and how it earns

Figures FY2025 (year ended December 31, 2025) unless noted; reported in JPY, USD at ~¥160/$. Primary line is 4004.T (Tokyo); a thin unsponsored OTC ADR trades as SHWDY.

What it is & how it earns. Resonac Holdings is a Japanese advanced-materials conglomerate formed by the 2020 takeover of Hitachi Chemical by Showa Denko and the 2023 rebrand to Resonac. FY2025 group revenue was ¥1,347.1B (~$8.4B, −3.2% on commodity-chemical drag), split between a high-value Semiconductor & Electronic Materials business — advanced back-end packaging materials, CMP slurries, copper-clad laminate, and top-3-globally EUV photoresist — and a large commodity base (petrochemicals, graphite electrodes, hard disks). The semi segment is ~38% of revenue but >99% of group core operating profit (¥506.3B revenue, +13.7%; record core OP ¥108.4B, +47%; AI-share of back-end materials ~10%→~20% YoY); everything else nets roughly to breakeven. Management's plan is to spin off the commodity petrochemicals (the new "Crasus Chemical" entity) and keep the semiconductor crown jewel — not to spin off the semi business. A large net-debt overhang (~¥681-707B / ~$4.3B) remains from the ¥980B Hitachi Chemical deal.

Cash-flow anatomy. The cash engine is the semi segment; the commodity tail drags margin and consumes maintenance capex:

The semiconductor-margin uplift (record core OP +47%) is offset at the FCF line by commodity-chemical losses, ~¥110B+ annual capex, and D&A of ~¥94B, keeping reported FCF thin despite the operating inflection. (FY2023-24 OCF/FCF are estimates; FY2025 capex/FCF are from the tanshin.)

Balance sheet & capital allocation. Net debt is significant at ~¥681-707B post-Hitachi-Chemical, deleveraging (gross interest-bearing liabilities ¥1,023.7B→¥969.5B in FY2025; equity ratio ~33%) via non-core divestitures whose impairments cut GAAP net income −60.5% to ¥29.0B. The dividend is ¥65/share (~40% payout); buybacks are minimal. The petrochemical (Crasus) separation is the SOTP / value-unlock catalyst — currently delayed pending Japanese tax/competitiveness legislation, so no shareholder vote yet.

What drives it. The bull case is the AI/semiconductor-materials re-rating (the stock already ~5-6×'d, ~40× forward) plus the petrochemical separation and deleveraging; the bear is commodity-chemical cyclicality, single-engine concentration, and AI-capex sensitivity. The question the tree resolves: after a 6× run, is a confirmed-but-richly-valued single-engine cyclical with ~¥681B net debt still worth owning, or is the easy money already priced in?

I. One-sentence verdict

Resonac is a CONFIRMED back-end AI-packaging-materials leader (CMP slurry, copper-clad laminates, advanced packaging materials) whose Semiconductor & Electronic Materials segment now generates >99% of group core operating profit and is compounding on AI demand (FY2025 segment +13.7%, record core OP +47%; Q1 FY2026 +21%/+74%; AI-share of back-end materials ~10%→~20%). The 2026-05-03 scaffold mis-framed this as a "speculative ABF second-source candidate" at ¥4,250 — but the thesis has already CONFIRMED and the stock has ~5-6×'d to ¥18,720/$113.68. The investment question is no longer "is the AI-materials story real?" (it is) but "is there still upside at ~40× forward, with a single profit engine, ~¥681B net debt, and a thin unsponsored ADR?" — and the honest answer is: don't chase; wait for a cyclical pullback.

II. The five facts that drive everything

FACT 0 (R2-VERIFIED, the new anchor) — the AI-back-end-materials engine is CONFIRMED and dominant. (Tier-A, FY2025 + Q1 FY2026 決算短信.)

The Semi & Electronic Materials segment is ~37.6% of revenue but >99% of group core operating profit — Resonac IS a back-end AI-packaging-materials company with a commodity-chemicals tail being divested. FY2026 guidance: revenue ¥1,310B, core OP ¥140B (+28%), EPS ¥425.45; H1 guide raised 40%, full-year maintained on Middle East/Hormuz uncertainty. The GAAP net collapse (−60.5%) is divestiture impairments, not operating weakness — which is why trailing P/E ~95× is a denominator artifact and forward is ~40×.

1. Resonac was created by the 2023 rebrand of Showa Denko KK + Showa Denko Materials (the latter formerly Hitachi Chemical, taken over by Showa Denko in 2020 for ~¥980B). The semiconductor + electronic materials business is the post-merger crown jewel; commodity petrochemicals are being divested. ✅A (now primary-confirmed: the segment generates >99% of core OP)
2. Resonac is the world's top-3 EUV photoresist supplier alongside JSR and Tokyo Ohka Kogyo (TOK). EUV photoresist is structurally oligopolistic; new entrants face high technical barriers. This is the strongest single moat in the Resonac portfolio. ✅B
3. Resonac competes with Ajinomoto's ABF in build-up dielectric film for FC-BGA substrates, but no public design-in announcements at named substrate-makers (Ibiden / Unimicron / Shinko / AT&S / Samsung SEMCO) for AI-accelerator-grade FC-BGA exist as of 2026-05-03. Market share is "single-digit" or low-teens at best in the advanced-FC-BGA segment vs Ajinomoto's >95%. ⚠️C
4. Pricing-power-circularity (per Codex 2026-05-03 AJNMY review) — if Ajinomoto pursues Palliser's +30% ABF price hike aggressively, substrate-makers have economic incentive to qualify Resonac as second-source. Industry-standard qualification timeline 18-36 months × end-customer sign-off 6-12 months = 3-5 years from trigger event to meaningful share displacement. ✅B
5. Hitachi Chemical goodwill (~¥600B+) on the balance sheet carries impairment risk if semi-electronic-materials segment growth underdelivers vs 2020 acquisition expectations. Plausible impairment scenario range: ¥30-100B. Not yet triggered. ⚠️C

III. The H-0 thesis

H-0 (R2-REFRAMED, one sentence): Resonac is a confirmed back-end AI-packaging-materials leader whose Semiconductor & Electronic Materials segment (>99% of group core OP, +47% record FY2025, AI-share ~10%→~20%) has already driven the equity through a ~5-6× re-rating to ¥18,720/$113.68 — so the structural-blindness mispricing the scaffold posited has largely CLOSED; what remains is a confirmed-but-richly-valued (~40× forward) single-engine cyclical with a leveraged balance sheet, where the forward edge is modest and gated on AI-capex durability rather than on the market "discovering" a hidden thesis.

Mispricing taxonomy (R2-updated): The scaffold's "structural-blindness on second-source emergence" was largely correct directionally — but it has already been resolved by the market: the 6× run is the re-rating. The residual situation is not a mispricing to exploit but a valuation question: at ~40× forward, the AI-materials growth is substantially priced in. Three layers:

1. The back-end AI-packaging-materials engine (CMP slurry, CCL, advanced packaging) — CONFIRMED, the real driver, now the >99%-of-core-OP fact. Re-rating largely complete.
2. The EUV photoresist oligopoly (with JSR + TOK) — a durable secondary moat that supports the floor.
3. The ABF-substitute optionality vs Ajinomoto — still speculative, still no public flagship-FC-BGA qualification (per update_2026-05-25.md); now an upside option, not the core thesis.

5 falsification conditions (Bull-thesis-breakers):

• FF1: Resonac EUV photoresist segment loses to a 4th competitor by FY28 (Korean / Chinese new entrant scales); oligopoly economics compress
• FF2: Hitachi Chemical goodwill impairment >¥50B announced (capital-allocation track record breaks)
• FF3: Ajinomoto monetizes ABF without triggering second-source qualification (the pricing-power-circularity argument breaks both ways — bad for Resonac if Ajinomoto succeeds in monetization without losing share)
• FF4: NAND/AI-semiconductor capex pause arrives 2026 H2 / 2027 H1 (same factor as TOTDY); Resonac segment growth compresses
• FF5: TSMC CoPoS panel-level glass eliminates organic build-up dielectric entirely by 2028 (extreme glass scenario; would hurt both Ajinomoto AND Resonac, not just Ajinomoto)

IV. EUV photoresist oligopoly (L1A — primary moat)

The single most durable moat in the Resonac portfolio. Treated as L1A precisely because it's underweighted in the prevailing narrative.

Leaf 1.1 — Top-3 EUV oligopoly is structural. JSR + TOK + Resonac collectively hold ~95%+ of EUV-grade photoresist supply. Each leading-edge fab (TSMC N2, Samsung 2nm, Intel 18A) requires more EUV photoresist as feature sizes shrink and exposure-counts increase. New entrants face 5+ year technical-development barriers. Verdict: ✅B — supported (Tier B).

Leaf 1.2 — Resonac's pricing power within the oligopoly is real but bounded. All 3 suppliers (JSR + TOK + Resonac) maintain pricing discipline historically; commodity-cycle behavior is dampened compared to substrate dielectrics. But pricing power is bounded by customer concentration (TSMC + Samsung + Intel = 90%+ of EUV demand). Verdict: ⚠️B — partial.

Leaf 1.3 — EUV photoresist segment growth tracks fab capex 1:1+. Each new EUV layer requires more photoresist; FY26-FY28 EUV layer-count expansion is multi-percent each year on top of unit-volume growth. Verdict: ✅C — supported but Tier C because the specific layer-count math hasn't been primary-pulled.

L1A roll-up: 2 ✅ + 1 ⚠️ — moat is real and durable but Tier-C-heavy on the specifics.

V. ABF substitute / second-source candidate position (L1B — speculative upside)

This is the narrative-friendly story but carries weaker evidence. Treated as L1B (not L1A) deliberately.

Leaf 2.1 — Resonac's back-end packaging materials are winning at record scale. [R2-UPGRADED ⚠️C→✅A] The primary FY2025/Q1 data confirms back-end semiconductor materials at a record quarterly high on advanced/AI demand (segment +13.7% FY2025, +21% Q1, AI-share ~10%→~20%). The broad back-end-materials portfolio (CMP slurry, CCL, packaging) is demonstrably winning. Caveat: the specific advanced-FC-BGA dielectric flagship qualification (the direct ABF-substitute) is still NOT publicly named — so the ABF-substitute optionality (Leaf 2.2) stays ⚠️, but the engine-level product success is now Tier-A confirmed. Verdict: ✅A — supported (engine confirmed; ABF-flagship-specific upside remains separate).

Leaf 2.2 — No public design-in at named AI-accelerator FC-BGA substrate-makers. As of 2026-05-03, no announcement that Ibiden / Unimicron / Shinko / AT&S / Samsung SEMCO has qualified Resonac MCL-E-679-G (or successor) for an AI-accelerator-grade substrate. Absence of evidence is NOT evidence of absence — qualifications are systematically opaque — but the market has no public reason to price Resonac AS IF substitution is happening. Verdict: ⚠️C — partial.

Leaf 2.3 — Substrate-maker economic incentive to qualify second-source IS real. Per Codex's pricing-power-circularity analysis (AJNMY review 2026-05-03): Ibiden / Unimicron / Shinko / AT&S have margin pressure from CoWoS supply tightness; +30% ABF price hike (Palliser demand) would compress their margins further; they have strong reason to qualify alternatives even if qualification takes 3+ years. Verdict: ✅B — supported (Tier B).

L1B roll-up (R2-updated): 2 ✅ + 1 ⚠️ — the back-end-materials engine is confirmed winning at record scale (2.1 ✅A); the specific ABF-flagship substitution (2.2) remains the speculative upside option. The substrate-maker incentive (2.3) is now corroborated by Ajinomoto's actual 30% ABF price hike (confirmed 2026-05-13).

VI. Capital allocation + Hitachi Chemical goodwill (L1C)

Leaf 3.1 — Strategic transformation toward semi-electronics is stated and underway. Multi-year IR commentary; capacity expansion at multiple Japanese sites; commodity-petrochemicals divestiture progressing. Verdict: ✅B — supported.

Leaf 3.2 — Hitachi Chemical goodwill impairment risk. ~¥600B+ goodwill on balance sheet. If semi-electronics-materials segment underdelivers vs 2020 acquisition expectations, impairment is plausible. Range estimate: ¥30-100B. Not yet triggered, but is the marginal fatal-flag candidate. Verdict: ⚠️C — partial.

Leaf 3.3 — Capital return discipline is modest, not load-bearing. Dividends maintained through cycles; minimal buybacks historically. Capital deployed to debt paydown post-Hitachi-Chemical acquisition. Not a leveraged capital-return story. Verdict: ⚠️B — partial.

L1C roll-up: 1 ✅ + 2 ⚠️.

VII. Mobility / EV battery materials adjacency (L1D — secondary upside)

Leaf 4.1 — Resonac has a Mobility segment with lithium-ion battery electrode binders, electrolytes, and electronic components for EVs. Less discussed in IR materials than semi-electronics. ⚠️C — partial visibility.

Leaf 4.2 — Demand exposure to China BYD-tier EV scaling is asymmetric: if China EV demand sustains, Mobility segment becomes meaningful upside; if China EV demand reverts to slow growth, Mobility segment is neutral. ⚠️C.

L1D roll-up: 0 ✅ + 2 ⚠️ — under-discussed optionality, not load-bearing.

VIII. Macro factor exposure (L1E)

Leaf 5.1 — Same NAND/AI-semiconductor capex factor as AJNMY/TOTDY/NVDA. If AI capex normalizes, Resonac's segment growth compresses. Same systematic risk as the rest of the AI-sleeve. Verdict: ⚠️B.

Leaf 5.2 — JPY exposure for USD-funded investors. Yen translation is meaningful for any USD-denominated holder; BOJ normalization or yen rebound amplifies returns. Verdict: ⚠️B.

L1E roll-up: 0 ✅ + 2 ⚠️.

IX. Three valuation scenarios

(See scenarios.md for the full analytical breakdown.)

R2-RE-ANCHORED 2026-05-30 at $113.68 (was ¥4,250 — a +340% under-anchor). The scaffold's ¥4,250 targets are obsolete by ~4×; the stock has already re-rated through them. At ~40× forward, the scenarios now hinge on whether the AI-materials multiple holds or compresses toward a cyclical-materials average — not on the market "discovering" the thesis.

Probability-weighted expected return: ~−2% (12 months). Asymmetry ~0.74:1 UNFAVORABLE (bull +28% vs bear −38%) — the inverse of a value entry. After a 5-6× run to ~40× forward, the risk/reward for a new position is poor: you're paying a growth multiple on a cyclical materials company with a single profit engine and a leveraged balance sheet, where the bear (multiple compression + AI-capex sensitivity) is larger than the bull. This is the analytical basis for "don't chase" — the thesis is right, but the entry is late.

X. Triggers and red flags

Triggers (Bull-case):

• T1 (anytime): Substrate-maker public disclosure of Resonac qualification at AI-accelerator-grade fab line — strongest single signal
• T2 (FY27-FY28): Ajinomoto pursues +30% ABF price hike and substrate-makers respond by accelerating second-source qualification — 18-month leading indicator
• T3 (anytime): EUV photoresist segment volume growth surprises upside (>20% YoY)
• T4 (anytime): China EV demand sustains; Mobility segment becomes meaningful upside

Red flags (Bear-case):

• RF1 (anytime): Hitachi Chemical goodwill impairment ≥¥50B announced
• RF2 (FY27 H1): NAND/AI-semiconductor capex cycle pauses; Resonac segment compresses
• RF3 (FY28-FY29): EUV photoresist 4th-supplier entry confirmed; oligopoly economics compress
• RF4 (extreme scenario): TSMC CoPoS panel-level glass eliminates organic build-up dielectric entirely by 2028

XI. Long-term holdability verdict

Per durability_test.md: aggregate score ~19/25 (R2-revised up from 17/25 — the primary-source confirmation of the back-end-materials engine strengthens business-model persistence + moat trajectory; a full durability re-run is owed but the direction is clearly up). 0 fatal flags, but the now-deleveraging Hitachi-Chemical goodwill + the single-engine concentration remain the marginal caveats.

Position-sizing recommendation (R2-UPDATED)

The scaffold's gating constraint was evidence quality ("Tier-C, get primary sources before sizing"). R2 RESOLVES that — the financials are now Tier-A and the thesis is confirmed. But a new constraint replaces it: VALUATION. After the ~5-6× run to ~40× forward, the risk/reward for a new position is unfavorable (EV ~−2%, asymmetry ~0.74:1).

• Verdict: Watch / Hold-small — do NOT chase at $113.68 / ~40× forward. This is a confirmed-quality AI-materials leader whose re-rating you (and the scaffold) missed; the disciplined action is to wait for a cyclical pullback (AI-capex air-pocket, a maintenance-shutdown quarter, or a multiple de-rate toward 25-30× forward) for a better entry.
• If initiating anyway (for structural AI-back-end-materials exposure): 1-2% maximum, sized small for the single-engine concentration + ~¥681B net debt + thin unsponsored SHWDY ADR liquidity (a real execution constraint — the TSE 4004 line is the liquid market).
• Hard cap: 2% — the combination of cyclical sensitivity, valuation, single-engine concentration, and ADR illiquidity does not support more.

FOR AJNMY THREAT-MONITORING:

• No position required. Track substrate-maker disclosure quarterly via triggers_redflags.md watchlist.
• AJNMY trim signal: if T1 fires (substrate-maker public disclosure of Resonac qualification at AI-accelerator-grade FC-BGA), trim AJNMY position regardless of own Resonac action.

Concentration-risk note (per K.3.4)

This position recommendation assumes the holder's existing AI/semiconductor sleeve is <25% of total portfolio. If existing AI exposure is heavier (e.g., owner's stated 45% AI concentration), Resonac adds correlated factor exposure — would NOT diversify the AI-substrate thesis but would concentrate it.

Owning both AJNMY and Resonac is the worst pair-trade structure (factor concentration without diversification). The Codex-aligned theoretical pair trade is "Long Resonac + Short AJNMY" — bets that aggressive Ajinomoto pricing fires and substrate-makers qualify Resonac. Most retail investors avoid shorting; "Long AJNMY only" remains the simplest expression of the bull thesis.

XII. Investment Scorecard (15-question quick reference)

Following MANUAL_en.md Part K.6 + K.10 (AI-sleeve archetype reading guide). K.3.5 weighted-score framework applied below.

Scorecard summary

K.3.5 Weighted-score derivation

Applying the 4-tier weighting from MANUAL §K.3.5. R2-RECOMPUTED — the primary-source confirmation of the back-end-materials engine flips the business questions from Tier-C guesses to Tier-A facts and lifts several from ⚠️→✅:

73% = moderate buy-with-sizing band per K.3.5 — a +13pp upgrade from the scaffold's 60%, reflecting that the business is now Tier-A confirmed (not Tier-C speculation). NOTE the divergence: structural quality is now solidly "moderate buy" (xii 73%), but the price-level verdict is still "don't chase" — the scorecard rates the business, while the §XI/§IX valuation work rates the entry. At ~40× forward post a 6× run, a high-quality business can still be a poor entry.

Final verdict

WATCH / HOLD-SMALL — do NOT chase. [R2-UPDATED] The scaffold's "WATCH only — insufficient evidence" reason is resolved: the financials are now Tier-A and the AI-back-end-materials thesis is CONFIRMED (xii 60→73%, h0 55→68%, archetype de-fragilized). But the stock already re-rated ~5-6× to ~40× forward, so the disciplined verdict is now valuation-gated:

• Don't initiate at $113.68 — EV ~−2%, asymmetry ~0.74:1 unfavorable. The re-rating you'd be buying has already happened.
• Wait for a cyclical pullback (AI-capex air-pocket, a maintenance-shutdown quarter, a multiple de-rate toward 25-30× forward) for a better entry into a genuinely good business.
• If holding for structural AI-materials exposure: 1-2% max, sized for single-engine concentration + ~¥681B net debt + thin unsponsored SHWDY ADR liquidity.
• Watch triggers: FY2026 results (~May 2027) for whether H2 OP exceeds the maintained ¥77B net guide; any named-substrate-maker FC-BGA qualification (the ABF-substitute optionality); AI-capex commentary for the cyclical tell.

Pre-purchase Section XII checklist: evidence quality is now Tier-A on the load-bearing financials (the scaffold's Tier-C gate is cleared). The remaining constraint is price, not proof.

XIII. What's NOT in this tree (deferred / documented but not built)

• Stage 1-3 supporting files: developments.md, consensus.md, assumptions.md, mispricing.md, taxonomy.md, frameworks.md, questions.md, leaves.md (full leaf decomposition)
• Stage 4 supporting: peers.md (peer comp), implied_prob.md (reverse-engineered probability)
• Stage 5 supporting: triggers_redflags.md (formal trigger doc), dashboard.md (live state)
• Stage 7 bilingual: tree_v1_zh.md (no Chinese companion yet)
• Evidence JSONL: no evidence_2026-05-03.jsonl — no Tier-rated source citations pulled
• External research packets: no chatgpt_review_packet.md or codex_review_packet.md for Resonac specifically

These are deferred to a full Stage 0-7 Resonac tree if owner upgrades from threat-monitoring to active position consideration.

Last updated 2026-05-30 (R2 re-anchor + thesis reframe). Source quality: Tier-A on the load-bearing financials (FY2025 + Q1 FY2026 決算短信 from Resonac IR) — the scaffold's "Tier-C throughout" gate is cleared. This R2 promoted RESONAC from a Tier-C "AJNMY-threat sketch" to a primary-source standalone investment memo: re-anchored ¥4,250→$113.68 (+340% under-anchor corrected), reframed the thesis (back-end AI-packaging-materials engine confirmed, not speculative ABF-substitute), h0 55→68%, xii 60→73%, verdict Watch/Hold-small (don't chase at ~40× forward post a 6× run). See evidence_2026-05-30.jsonl + update_2026-05-30.md. The remaining ⚠️ leaves (ABF-flagship qualification, Mobility, macro/cyclical) are genuinely unresolved, not data-gaps. A full Stage 0-7 durability re-run is owed.