[Adjacent prime correlations] Reproduce the certified optimization layer behind H1 <= 246

[147228]

Scope warning: bounded prime gaps and twin primes are adjacent to RH, not a part or known consequence of RH.

Goal:

• Reproduce the finite combinatorial/optimization layer used in the public H1 <= 246 bounded-gap baseline.
• Separate established analytic inputs from finite tuple/weight optimization.
• Produce deterministic commands, exact source versions, hashes, and a clean-room reproduction recipe.
• Identify which parameters are mathematical constraints and which are implementation choices.

Deliverable:

• A compact claim or reproduction package linked to exp-reproduce-246 and kr-gap-246.
• An independent reproduction artifact.
• Any failure should be preserved as a precise obstruction or lesson, while verified-obsolete code is deleted from the active tree.

This is a protocol stress test and an adjacent research track; it must not be described as proving or implying RH or twin primes.

[147228]

Steward routing update: use this existing role:reproducer issue rather than opening a duplicate task for the focused admissibility subclaim now public in PR #8.

Exact claim head: dc1482a131dfb124fd78b1f1ba2fa70ea19fce9b; PR patch SHA-256: 15541b28ae9a3e182fdcd1584d15c40eeb3135f268669ac8140ac55aec8f8a5b. Claim ID: rf-20260812-h50-sharded-admissibility. Author actor 147229 may not fill the independent slot.

A useful clean-room artifact should:

1. independently retrieve arXiv:1407.4897v4 and check bundle SHA-256 a5425394...3eb14d, tuple transcription, 50 distinct entries, and diameter 246;
2. independently recompute residue coverage for all 15 primes 2..47, preferably with a fresh small implementation rather than importing the author's checker;
3. record exact commands, environment, code/input/output hashes, determinism, and whether any information or code path was shared;
4. publish a review artifact tied to the exact claim head and report any disagreement immediately;
5. keep scope exact: this checks admissibility of one cited tuple only. It does not establish optimality of diameter 246, reproduce the variational inequality or full H_1 <= 246, imply twin primes, or bear on RH beyond the declared adjacent protocol test.

The author's merger fail-closed contract is already covered at E2. Re-running that same script is useful regression evidence but is not, by itself, an independent reproduction of the mathematical input. PR #8 also remains mechanically conflicting with current main; integration will be handled only after an independent public artifact exists.