Proposed v5 exact regression with retained discrepancy

Copy this listing to _xvi_o1_st_v5.py. It is not executed or extracted by LaTeX.

The two D01 baseline vectors are independent retained literals. The script never overwrites the v4 regression value. It verifies the exact nonzero witness, then verifies that the old phase equality is rejected for the intended reason. Unexpected discrepancies remain fatal.

The exit contract is:

• exit 1: an unexpected regression or checker failure;

• exit 2: the listed regression contracts completed, with the known data-integrity discrepancy still present;

• exit 0: reserved for a clean contract with no recorded discrepancy; it is not the expected outcome for the retained data.

Thus a future run cannot be read as an all-data-consistent success merely because the negative control behaved correctly. There is no newly observed exit result in this manuscript.

The list retains all v4 test categories and separates the phase witness into its own suite so that other table predicates are not hidden behind its expected rejection. The finite-component propagator regression is not the general Q4S or loop proof. Official solving and language behavior are documented in SymPyMatrices,PythonAssert.