Observable, readout, and statistical error ledgers

Let the implemented source coefficient be $\hat{a}$. Use

for its two displaced states. If a scattering or packet calculation supplies a separate weighted coefficient error $r_{\mathrm{sc}}$, include it in the target comparison using Lemma lem:weylcontinuity.

A sufficient per-preparation observable allowance is

The terms respectively concern source-profile error, certified coefficient error, compact packet approximation, local transfer, probe noise, and final readout.

A final readout allowance must be justified for a complete common record alphabet. Ordinary binary outcomes and all declared reset, assignment, and missing-record flags are retained. Assign every failure a fixed score in $[0,1]$; no trial is removed. A complete-law total-variation error changes such a score expectation by at most that error.