Why zero simulated bias is not physical calibration

Setting $b_{\mathrm{cal}}=b=0$ inside a simulator does not exclude the second physical completion. This proves the requested calibration non-transfer rather than merely restating it.

A memory-channel certificate does not estimate $S_{\Omega }(A)$, verify the local Hamiltonian, prove uniqueness, or certify the full-system gap. Those require their own state, spectral, and measurement interfaces.