FROSTYVERSE H2-083
Natural Four from Two Masses Search

PURPOSE
-------
The known Frostyverse bookkeeping is approximately:

    stable FV ~= +2D
    missing   ~= +4D
    total     ~= +6D

The exact H2-083 question is NOT whether a small mass correction can be added
to a coefficient that is already set to 4.  Instead it asks whether the 4
itself can emerge from a dimensionless relationship between the two masses:

    C = F(M1, M2)

    total = stable FV + C*D

where M1 is the source mass (Sun in named tests) and M2 is the orbiting body.

STRICT RULE
-----------
Candidate formulas contain NO literal target 4, NO 4.0498328559103465, NO
Mercury-fitted amplitude, NO fitted exponent, and NO per-planet normalization.
The desired C=4 appears only as the external target used to rank the Mercury
discovery candidates.

WHY MASS RATIOS
---------------
C is dimensionless.  With only two masses and no new dimensional scale, the
scale-free information is a function of their ratio (or algebraically
equivalent total/reduced/symmetric mass combinations).  The search therefore
uses M1/M2 and canonical combinations constructed only from M1 and M2.

SEARCH SERIES
-------------
1. Seal a coefficient-free discrete expression grammar before holdouts.
2. Search Sun+Mercury only for formulas naturally landing within 5% of C=4.
3. Freeze every surviving nondegenerate Mercury formula.
4. Quarantine algebraic/numerical constant identities.  Example of the kind of
   trap this prevents: log ratios of nested roots can equal an integer for
   every possible mass ratio.  Such a formula has not learned anything from
   the masses even if it equals exactly 4.
5. Open Sun+Venus, Sun+Earth, and Sun+Mars as named holdouts.
6. Test a synthetic mass ladder from 0.001x to 10,000x Mercury mass.
7. Test a generic planet/star mass-ratio grid q=M2/M1 from 1e-10 to 1e-3.
8. Rebuild the inherited FV stable+D ledger at R20/R40/R80 and apply each
   frozen mass equation to the actual total response.
9. Rank the frozen equations only after all holdouts are opened.

EXPRESSION GRAMMAR
------------------
Base:
    M1/M2

Unary operators, depth <= 3:
    ln(x)
    sqrt(x)
    cbrt(x)
    1/x

One binary combination of generated terms:
    +, -, *, /

A small canonical family of direct two-mass sum/product/reduced-mass
combinations is included as well.  No target constant or fitted continuous
parameter is introduced.

DISCOVERY / HOLDOUT FIREWALL
----------------------------
Sun+Mercury plus the external desired number 4 is the only discovery target.
Venus, Earth, Mars, synthetic mass ratios, and the FV response ledger are not
used to select the frozen Mercury candidate set.

IDENTITY / TAUTOLOGY FIREWALL
-----------------------------
Every candidate is evaluated on a wide generic q grid before promotion.  If
its value is constant to numerical precision across that grid, it is labeled
DEGENERATE_CONSTANT_EQUIVALENT and cannot count as evidence that two masses
produce the coefficient.  This is essential because algebra can trivially
hide fixed integers inside roots/logarithms while displaying mass symbols.

SCIENTIFIC STATUS
-----------------
This is a TARGET-AWARE empirical equation search.  A successful formula would
be a calculable two-mass closure candidate for the missing ~4D coefficient.
It would still not explain the microscopic FrostTransit mechanism that chooses
that formula.

No G change, thrust, kick, Sun spin, stronger ordinary gravity, free energy,
or master update-law change is made.

MASS CONSTANTS
--------------
Fixed JPL DE440 GM values inherited from H2-082 are used.  GM ratios equal the
corresponding mass ratios because G cancels.

OUTPUTS
-------
Main report at package root:
    h2-083results.txt

Detailed CSV directory:
    H2-083_CSV/

Key CSVs:
    h2-083_mercury_discovery_frozen.csv
    h2-083_degenerate_constant_identity_audit.csv
    h2-083_named_mass_holdout.csv
    h2-083_synthetic_mass_holdout.csv
    h2-083_generic_q_stress.csv
    h2-083_stable_D_target_ledger.csv
    h2-083_named_FV_ledger_predictions.csv
    h2-083_synthetic_orbit_mass_FV_predictions.csv
    h2-083_blind_holdout_ranking.csv

BASELINE
--------
Built from Frostyverse-Test-H2-082.zip
SHA-256:
    8e29faf7bf0b7fca468ef6cb2afe5dfa4978b317b7cd1284fd162a72459c3e77

Only the H2-083 launcher/runner/readmes are at package root.  The inherited
engine dependency set is unchanged.
