The preregistered small arithmetic survey had zero post-top black vertices in its admissible tested range. It therefore passed no actual blackness witness and failed the nonempty assertion. This was a test-coverage failure. The independent lift-count, Markov and exact clock controls ran before this assertion. Keep the empty survey explicitly and add constructed vertices using 2^d = 1 mod 3^k. These fixtures are deliberately chosen and do not estimate how often such vertices occur.
The future-lift checker initially required more than 100 retained comparisons. There are exactly three eligible four-bit prefixes, 1111, 1110 and 1101, and 31 residue/precision rows per prefix. The measured 93 successful comparisons caused the overly high coverage assertion to fail. The revised assertion checks the exact expected count. No checked congruence or direct parity reconstruction failed.
The first finite late-clock proxy at gap 32 with p=2/3 hits equality of the mean and the threshold after flooring, contrary to an overly strong test assertion. The proof uses strict separation only for sufficiently large gaps and absorbs the small cases into a constant. The corrected check retains this equality row and checks strict separation from gap 64 onward in the two specified proxies.
