STONEHENGE: COMMON-EPOCH DOMINANCE AUDIT — 30 October 2026

This is a new analysis note, not a replacement manuscript or submission package.

RESULT
For every Cygnus epoch strictly inside approximately 3051.2428–2927.2670 BCE,
its fixed-scale RMS and reference-radius RMS are both below every rival's
separately optimized minimum in the complete 19,939-field nominal census.
No horizon or four-arm filter is needed for this two-objective statement.

The dynamic-radius objective also satisfies this comparison after excluding
the one field already demonstrated to remain invisible from Stonehenge
throughout the entire search. All Cygnus stars satisfy the necessary horizon
condition over the common interval. Minima of all remaining competitors are
unrestricted lower bounds: restricting them to locally feasible epochs can
only increase their errors. This proves the three-objective version for the
necessary-horizon candidate family without claiming a new feasible count.

At the common example epoch 2985 BCE:
Criterion              Cygnus          Best rival minimum (lower bound)
Fixed RMS (degrees)    0.1017542495     0.2146951964
Reference-radius RMS  0.0066372249     0.0087166264
Dynamic-radius RMS    0.0061610046     0.0086934008 [horizon scope above]

PROPOSITION
Let f_jk(t) be objective k for candidate j and let
b_k = min over all rivals j and all allowed t of f_jk(t).
If f_Ck(t0) < b_k for every k, then for any weights w_k >= 0, not all zero,
and arbitrary positive normalization constants a_k,
  sum_k w_k a_k f_Ck(t0) < sum_k w_k a_k f_jk(t)
for every rival j and every allowed epoch t. The result follows by multiplying
each component inequality by w_k a_k and summing. Any strictly increasing
aggregation also preserves componentwise dominance. Thus the winning stellar
field cannot be changed by reweighting these specified objectives.

This is stronger than a Pareto-front membership statement: a common Cygnus
epoch beats lower bounds on every competing field in every listed objective.
The proof is algebraic; its premises are numerical minima from the archived
grid-plus-refinement search, not interval-certified global optimization.

SCOPE
The interval is a numerical dominance region, not a confidence interval or an
archaeological date estimate. The objectives share data and are not independent
confirmations. This adds robustness to criterion weighting, not evidence that
prehistoric builders intentionally encoded the stars. It does not cover changed
count maps, free permutations, arbitrary angular scales or new candidate sets.
Unrestricted free-ratio E still places Cygnus 68th in the complete family.
In the separate 283-field four-arm family, all four listed objectives share a
smaller common dominance interval, approximately 2989.2793–2981.6808 BCE.

The exploratory calibration-window calculation retains a single fixed Cygnus
epoch and gives a conservative scale interval 0.98495225 < s < 1.01441656 for
target s*(7,18,4). It is not the full optimized winning interval; it does not
justify arbitrary-scale robustness.

REPRODUCTION
Requires Python 3, NumPy and SciPy (executed with NumPy 2.3.5, SciPy 1.17.0).
From this directory:
  python joint_dominance.py
  python calibration_window.py 1
  python calibration_window.py 0.5
The two latter commands generate large, disposable RA caches (about 385 and
770 MB). They are not required to reproduce the joint-dominance result.
S8_excerpt retains unchanged catalogue, astrometric engine and complete score
tables from the supplied v1.41 research package. The dominance script reuses
those rival minima and recomputes Cygnus over the full epoch window. It does
not rerun the full optimization of all 19,939 competitors. Verification also
compared the new scores with the coordinate-free pole-vector formula.
