# Cross-Track 48h Anomaly Watch

**Date:** 2026-09-05 ~00:25 America/New_York  
**Window:** ~2026-09-03 → 2026-09-05 (plus edge items that still matter for reopen logic)  
**Question:** Does any new abs **force protocol reopen** on Prop-01a / 02a / 03?  
**Rule:** Real abs only (HTTP 200). No fabrication. Conservative bar: reopen only if it breaks locked success criteria, invalidates E1/Magpie-lite design, or requires immediate Cobaya/stack swap.

---

## Verdict (all tracks)

| Protocol | Force reopen now? | Why |
|----------|-------------------|-----|
| **Prop-01a** (PhySO/E1) | **No** | No new units-SR engine; InsightSR still parked; GPU constant-opt is post-E1 optional only |
| **Prop-02a** (Magpie-lite / UQ-AL) | **No** | No new OC20/materialsUQ UQ paper; HT/device papers don’t change Magpie success flag |
| **Prop-03** (Cobaya / DESI / EDE) | **No force reopen** — **priority widen only** | NMC/sirens + Shark JWST-null affect *menu/weighting*, not locked Cobaya install path |

**Bottom line:** Keep E1 frozen, Magpie-lite ablation path, Cobaya install-first. Lit anomalies → **queue / widen**, not protocol reset.

---

## Anomalies reviewed (ranked by reopen risk)

### A1. NMC evolving DE + GW/EM sirens — `2609.04112` (Prop-03)
- Abs: https://arxiv.org/abs/2609.04112 · HTTP 200 · 2026-09-03 · Lagos & Wolf  
- **Anomaly?** Extends beyond axion-EDE framing toward NMC + \(\alpha_M(z)\) siren tests under DESI-like DE.  
- **Reopen Prop-03?** **No.** Does not invalidate Planck+DESI+SN Cobaya baseline.  
- **Action:** Already on 48h watch — **widen theory menu** after Cobaya stands up (priority widen, not reopen).

### A2. Shark: bright \(z>10\) galaxies in ΛCDM SAM — `2609.03537` (Prop-03 adjacent)
- Abs: https://arxiv.org/abs/2609.03537 · HTTP 200 · 2026-09-03  
- **Anomaly?** Weakens JWST-as-EDE-motivation narratives.  
- **Reopen?** **No.** Changes *rhetoric/weighting*, not likelihood stack.  
- **Action:** Keep JWST out of primary EDE justification (already flagged).

### A3. GPU constant-opt for tree GP SR — `2609.03352` (Prop-01)
- Abs: https://arxiv.org/abs/2609.03352 · HTTP 200 · 2026-09-03  
- **Anomaly?** Relevant if PySR constant-fitting is the E1 failure mode.  
- **Reopen Prop-01a / freeze E1?** **No.** Post-E1 optional only (CoS accepted).  
- **Action:** File for after measured E1; do not touch `e1_live.yaml`.

### A4. InsightSR LLM+PySR — `2608.25291` (Prop-01, edge >48h)
- Abs: https://arxiv.org/abs/2608.25291 · HTTP 200 · 2026-08-26  
- **Reopen?** **No.** Parked until PhySO/PySR rates measured (`lit-insightSR-vs-physo.md`).  
- **Action:** None until E1 artifacts.

### A5. HT 2D inverted resistive switching discovery — `2609.03578` (Prop-02)
- Abs: https://arxiv.org/abs/2609.03578 · HTTP 200 · 2026-09-03  
- **Anomaly?** HT discovery narrative, not Magpie-lite UQ metric.  
- **Reopen Prop-02a?** **No.** Different task (device RS ≠ composition/structure AL recall).  
- **Action:** Watch only; Writing Bot narrative later if needed.

### A6. EuIn₂Sb₂ axion insulator candidate — `2609.02114` (Prop-02)
- Abs: https://arxiv.org/abs/2609.02114 · HTTP 200 · 2026-09-02 (edge)  
- **Reopen?** **No.** Experimental target class, not AL protocol.  

### A7. Dark-sector × dynamical DE + DESI DR2 — `2609.03062` (Prop-03 edge)
- Abs: https://arxiv.org/abs/2609.03062 · HTTP 200 · 2026-09-02  
- **Reopen?** **No.** Strengthens joint-model queue already planned; does not replace Cobaya install.  

### A8. Cleared GPAge mid-time — `2608.28260` (Prop-03)
- Abs: https://arxiv.org/abs/2608.28260 · cite OK after PDF check  
- **Reopen?** **No.** Cobaya **reproduction still required** before any “tension resolved” language (unchanged CoS rule).

---

## What would force a reopen (trigger sheet)

Hold this sheet for next continuous watches:

| Track | Force-reopen trigger (examples) |
|-------|----------------------------------|
| 01a | Peer result showing PhySO units grammar *invalid* on Feynman protocol we use; or E1 design bug exposed by literature (matched-budget mismatch) |
| 02a | New standard UQ/AL benchmark that makes Magpie-lite metrics non-comparable / leaked by construction |
| 03 | Official DESI/Planck likelihood deprecation or Cobaya module break; or community consensus killing EDE *and* our DESI joint path simultaneously |

None of A1–A8 meet these triggers in this window.

---

## Sources
- Prior verified watches: `lit-prop01-7d-watch.md`, `lit-prop02-7d-watch.md`, `lit-prop03-48h-watch.md`, `lit-insightSR-vs-physo.md`  
- Abs HTTP 200 re-checked 2026-09-05 for A1–A7 ids  
- Fresh arXiv API partially rate-limited this pull; anomaly set drawn from already-verified 48h harvests