Bind pooled R437 replica-indexed native requests and charge widened source codes

[Clear1nsight]

PR3995 showed that the actual pooled R437 source cannot enter the retained replica-blind catalogue. This PR constructs the next missing interface: a full replica-indexed request table and native owner binding for all 903456 paid sites. Partial source/compiler result; physical CALL/RETURN and supplier admission remain open.

Every request key resolves back to its actual caller, including pooled absolute families and exactly-once IDLE enumeration. The explicit consumer API is idle_identifiers(child, replica); the legacy IDLE table is unavailable. COPY retains the common private target and records its actual source chart. Fixed source handles reject omitted/stale/cross-prime/mismatched contexts, including an incorrect Specifications catalogue.

All 1235 actual charts have dyadic entries and determinant ±1. The all 288-call audit certifies 179802 same-family owner pairs by universally nonzero dyadic-unit entry witnesses, without replica symmetry. This yields separation for every odd prime and invertible common low; it is not an enumeration or execution of GL64. The full profile has 836928 MOVE, 2592 COPY and 63936 IDLE sites, rank mass 1808352, 400877 rank-2 requests and one rank-14 request. Ordered request digest: 29ad864c2a6a31ad15162b0165b4f047cf580d9f2c147aebcfccfd564ccbbc60.

The finite source-code vocabulary grows 32-fold, adding exactly five bits at every fixed q (q=3: 6507→6512). The proof retains 379584 routes and 507456 family scans per low class and charges +5 per raw field, +5h per h-step Program, plus resulting chronological parking and consumer costs. Unresolved occurrence counts, physical tariffs, native/child source acceptance and global reserve are explicit unknowns. Four actual retained q=3 source specifications compile; their good/fallback status is recorded. No physical residual body, whole-volume caller, new kappa or frontier change is claimed.

Validation: standard-library registered normal/optimized replay, source authentication, ten negative controls and exact-byte publication checks; detailed evidence in VALIDATION.md. Same-session AI cross-review independently re-counted the literal pooled sites/profile/owner pairs, found two context-boundary omissions, and checked their fixes. This is not external human or formal-kernel acceptance.

Credit Rohan Arun's #3959/3958/3939 for pooling and endpoint; ai-nanachi3647, Ali Romman1043, Sagar Thakrar3700/2423 and the R437 lineage;3995 for absolute-owner emission/catalogue counterexample;3980 for the dyadic checker;3983 for chronological resource accounting; Nick Petalas2989 for earlier owner/compiler correspondence. The existing3959 local follow-up already establishes the first helper's permutation transport/common plan. Canonical physical/source credit includes zxv, David Arias, Jose Manuel/jmbarrancoml, ywz65, alephnan, Adam Gall, Ross Richardson and all retained notices. Chris R/Clear1nsight, substantial OpenAI Codex assistance, Apache-2.0.

Explicit donor assignment refines independent-moment. Canonical main 20ecf28 / revision 6 refreshed unchanged before publication. Current #3995/3959 comments, live R437 replica search, parent live open-body checks and local overlap procedure were inspected; broader closed/private/Discussion coverage remains partial. Shared claim: #3967 (comment) . This package does not duplicate #3995's emitted-IR proof or claim its packing improvement.

Final documentation head 92daee4bf12255efe4ae0102eeb5f72794bf9c57 passes exact154-file package/input byte, declared source-history, metadata and whitespace checks. Scientific bytes are unchanged after the normal, optimized and read-only Git-free replays.