Preserve aligned pairing and its conditional 17*2^-63 witness

[Bortlesboat]

This preserves the aligned-pair refinement of the historical 2^-59 construction against current upstream. Its certificate, proof packet and original feature code are unchanged. The README places its standalone conditional target alongside the newer, stronger community construction, with the remaining independent review obligations explicit.

Aligning the global pair groups between common-point circuits removes 14,944 additions at h=50. Under the same retained analytic recipe and upstream theorem, the exact bit saving supports a conditional target of κ = 17·2⁻⁶³, 6.25% above the earlier paired witness target 2^-59.

For common point i, put its global pair partner i ^ 1 last in the local point order and leave all other complete pairs together. Relabel local inputs and outputs consistently. The original 49-point paired circuit is unchanged; the existing fixed-pair-star sharing now identifies more equal sums across common-point groups.

• Additions: 450394 → 435450; side roles: 509194 → 494250; all 58800 partial outputs retained.
• Exact counts: W=394839648000000, s=49354954232864000000, unchanged deficit D=1767136000000, and eta_b=23/642375000.
• Bit saving: a_b=305/10^11, using the existing rational logarithm enclosure. With the retained complex motif, stopped guard and Gaussian recipe, all 30 strict constraints and seven margins pass. The exact minimum is G=739738521/400000000000000000000000000 > 17*2^-63.

Includes an exact certificate, a short proof note/PDF, a construction and dependency audit, and an independent patch against the unchanged pinned upstream source. The previous paired witness and its generated patch remain reproducible. This changes the graph supplying the bit exponent and complements #1's parameter refinement.

Current-base validation: full make verify passes 224 tests and 21 independent patch checks under Linux/GCC. Every upstream validation target is retained, and all regenerated certificates and patches match their Git blobs exactly. New coverage includes a full-size expansion of the actual global DAG in 1,024-input chunks, independent of compressed provenance; small dense support checks; both frame directions; and dirty-scratch/bank-exchange checks. The proof note and complete patched manuscript compile.

The complete upstream multiplication theorem and retained analytic interfaces remain assumptions. No family optimum, dyadic 2^-58 target or practical speedup is claimed.

Focused mathematical review: the review packet (three-page PDF) separates the finite claim from the conditional running-time result, identifies the rank/restoration, fixed-tape recurrence and downstream proof obligations, and gives reproduction commands. External expert assessment of those obligations is requested and remains pending.

[Bortlesboat]

@Mister-Meeseeks, I've added a focused review packet for this refinement: Markdown packet and three-page PDF, pinned at 1c200af99d348de496d78274de3584339dcf02e3.

It separates the independently reproducible finite circuit from the conditional running-time consequence and maps three review obligations to exact source labels:

1. R1: Does the aligned sharing preserve the full forward/reverse frame transfer, arbitrary dirty-scratch restoration, every role's endpoint, and the stated rank sum, including the negative X-source correction?
2. R2: Does prop:power-interchange make exactly s calls on logical volume V/W, with uniform O(V) node overhead on a fixed number of tapes? The packet includes the existing parking and descriptor arguments, and asks that copying, head movement, spectator ranges and ancestor cleanup be accounted for. lem:chunk-swap must preserve the saving for unrestricted widths.
3. R3: Do the retained layer, stopped-guard, resampling and assembly interfaces justify the conditional 17*2^-63 target with the new parameters?

Could you review R1 and help identify an expert uninvolved in the original construction/audit to examine R2? A specific source label and counterexample or missing implication would be especially useful. The packet gives commands for the full-size coefficient oracle and certificate; passing those checks is not being offered as validation of the machine-time bound or the complete theorem.

[Bortlesboat]

The conflict is resolved against current upstream 0605a24. Full make verify under Ubuntu/GCC passes 224 tests, the focused suites and all 21 independent patch checks; every regenerated certificate and patch matches its Git blob exactly. The original circuit and review packet are preserved.

The new workflow run for 8bb07d5 is awaiting fork-workflow approval and has executed no jobs. Could a maintainer approve it so the refreshed contribution gets an upstream CI result?

[Mister-Meeseeks]

Integrated and retained on main in 377e2bd. Thank you @Bortlesboat for this contribution.

The complete original source tree at 8bb07d55f6898bc8ec919222cf7596e987e9b69f is preserved in the historical source package, with original notices and predecessor credit. It can be extracted and replayed independently. Historical release wrappers do not overwrite current pinned suppliers.

Reviewed scope: Aligned common-point pairing: full h50 support DAG; small dirty-state basis and both frame directions; exact historical 17*2^-63 witness, scoped ceiling, and regenerated upstream patch.

The batch review and per-file dispositions cover every original changed path. Local focused checks passed; regenerated artifacts match the submitted sources. Attribution is also recorded in CONTRIBUTORS.md. This is integration of historical research and finite evidence, not a new selected κ or full formal verification of inherited tape/analytic interfaces. Linux replay jobs are tracked separately.