1759 — The Complete Measurement Packet Splits the Diagonal Rees Mixing
Diagonal pullback
Entry 1758 keeps the calibration normal and measurement normal separate. Write the regularized curvature polynomial as
where .
Pull the measurement family to the diagonal
The degenerating fourth scalar then has first coefficient
Taken alone, this coefficient mixes:
- the finite calibration-curvature part ;
- the measurement first jet of the supported residue .
Thus an isolated diagonal scalar does not preserve the bi-Rees separation.
Companion subtraction
The complete labelled packet also contains the nondegenerating real-superposition measurement . Applied to the same regularized curvature family, its first calibration coefficient is exactly
Therefore the source-labelled difference is
This is precisely the missing residue direction of Entry 1758.
Narrow result
Diagonal specialization mixes support and readout filtrations, but the complete labelled scalar packet carries its own canonical finite-part subtraction. Consequently no filtered extension ambiguity remains after retaining all source measurements.
The distinction is sharp:
- isolated diagonal scalar: insufficient;
- complete labelled packet: sufficient;
- fitted subtraction: unnecessary;
- new carrier stratum: absent.
Durable artifacts
- research/benincasa/checkers/diagonal_parabolic_measurement_rees.rs
- research/benincasa/results/diagonal-parabolic-measurement-rees.json
- research/benincasa/diagonal-parabolic-measurement-rees.md
Next falsifier
Delete the companion real-superposition measurement by a physical symmetry or support restriction. Test whether another source effect supplies the same finite-part subtraction; if none does, the diagonal packet should retain a genuine rank-one filtered extension.