FROSTYVERSE PHOTON FLY-BY PYTHON TEST — EXPECTED REFERENCE OUTPUTS
=================================================================

LIGHTWEIGHT FOCUSED REFERENCE
-----------------------------
This reviewer-facing fingerprint uses the unchanged H-099 mechanics for one
case from the historical validation profile:

  resolution R                   = 11 cells per source radius
  requested impact parameter     = 4.85 source radii
  actual snapped impact          = 4.818181818181818 source radii
  orientation                    = XY (travel +X, impact +Y)
  path half-length               = 3.0 source radii

Expected values (ordinary floating-point/platform variation permitted):

  wire nRMSE                     = 8.55205019718325e-03
  maximum FV-GR path separation = 9.751939371748257e-09 model-distance units
  maximum separation / nominal cell
                                 = 1.0727133308923083e-07 cells
  FV signed bend                 = -1.2488312900311113e-01 arcsec
  post-seal GR signed bend       = -1.2312153841364273e-01 arcsec

The exact FV seal fingerprint may depend on NumPy/platform byte representation;
the numerical result is the reproducibility target.


ENDPOINT-BOOKKEEPING REGRESSION — REVIEWER WRAPPER 1.0.2
-------------------------------------------------------
These fresh-input cases previously exposed the rounded-endpoint/whole-wall mismatch.
With the 1.0.2 reviewer wrapper they must complete without a forward-wall-count abort:

  R17, requested b=2.73, orientation ZX, half=4.2
    actual b                      = 2.7058823529411766
    FV signed bend               ~= -2.972332582502e-01 arcsec
    exact-endpoint GR bend       ~= -2.939156593799e-01 arcsec
    relative bend difference     ~= +1.128759%

  R18, requested b=6.37, orientation YX, half=5.3
    actual b                      = 6.388888888888889
    FV signed bend               ~= -1.074962260694e-01 arcsec
    exact-endpoint GR bend       ~= -1.068724377150e-01 arcsec
    relative bend difference     ~= +0.583676%

No photon steering coefficient or exponent is retuned by this correction.
The original H-099/H-109 archives remain unchanged; this regression applies to
the reviewer custom/reference lane only.

UNCHANGED H-099 VALIDATION PROFILE
----------------------------------
Historical validation runs R=(9,11), b=(2.35,4.85), and two orientations.
Representative summary rows include:

  R9  b=2.35 wire ~= 1.2579% ; worst max separation ~= 4.132551e-08
  R9  b=4.85 wire ~= 1.0570% ; worst max separation ~= 1.176755e-08
  R11 b=2.35 wire ~= 1.0090% ; worst max separation ~= 3.404245e-08
  R11 b=4.85 wire ~= 0.8552% ; worst max separation ~= 9.751939e-09

The validation profile deliberately does NOT reopen the high-resolution freeze
decision. It should end with the decision that the validation profile completed
and the high-R H-099 freeze-safety audit remains unopened.

PRESERVED FULL H-099 CHECKPOINTS
--------------------------------
The archived full audit reports peak physical miss / nominal FrostCell:

  b=2.35 : 2.430601057e-06 cells
  b=4.85 : 4.031342048e-06 cells
  b=6.25 : 4.591972922e-06 cells

These are nonzero residuals. The claim is not point-for-point identity with GR.

UNCHANGED H-109 VALIDATION
--------------------------
The H-109 validation profile should report exact-provider identity to numerical
precision, identical route/wall hits, no memory mapping in the direct-read lane,
and a fixed 64 KiB explicit exact-field cache.
