Canonical CIP Weakness and Host Degeneracy in an Interacting Dark-Sector Two-Point Test

This paper tests whether the charge-uncompensated compensated-isocurvature channel identified in the preceding Type-2 dark-sector analysis is statistically identifiable in controlled linear two-point data. Using the production two-mode CLASS basis, public ACT DR6 × unWISE bandpower windows and covariance, tracer and redshift-distribution nuisance structure, and the certified Planck high-ell host geometry, the paper separates the direct CIP contribution from the broader host-consistent interaction response. At the canonical point the direct-CIP transfer derivative is already extremely small before nuisance projection, while the measurable background/adiabatic interaction tangent remains strongly degenerate with ordinary host directions. A finite gamma = 0 to 3 Planck secant confirms that the host degeneracy is not merely a differential artifact. The result is deliberately an identifiability statement: it does not claim a coupling exclusion, a preference for interaction, or a globally free-CIP bound.