Explicit nonpositive normalization

This is not a proof that the entire class lacks positive physical representatives. It is a proof that the class data do not authenticate positivity of a representative. The same distinction appears in Cheung's covariance audit CheungXIXD.

Because $\mathcal{F}U=0$, the modification (eq:bad) does not change the subsidiary defect considered in Section sec:subsidiary. Its compact-support-on-time-compacts property also means that it does not alter an already proved one-sided asymptotic tail at spatial infinity.