Search state 2026-09-19 (route 35, revision 10; the route's own record carries the day's queries: #660's design triage, #1028's calibration, #1029's registered run; #1029's search of 2026-09-18 found no source that tests whether a deterministic admissible-slot census predicts twin occupancy beyond its slot count, and the methodological items it names, multi-level block permutation (PMC4644991) and Winkler et al. 2020 (PMC7573815) for the refit-per-draw nuisance rule, are the controls reused here as R1/R2). The step here is the instrument extension #1029 pre-registered for revision 10: finer block scales, a spread gate applied before any statistic is read, a spread-bearing grain statistic, and the gap-local scale the route named as its weakest assumption. Nearest prior work for the new pieces. (1) Finer scales and the spread gate: #660 §3 proved the structural reason gmax is near-constant (the number of blocks attaining the tile maximum is at most the tile's maxgap occurrence count: 20 at x = 19, 4 at x = 23) and recommended declaring such cells vacuous and adding "the block's count of gaps in the tile's top decile of gap values"; that is the topdec statistic used here, with its threshold fixed from the whole period's gap distribution. (2) The gap-local test: occupancy of admissible slots conditioned on the size of the neighbouring admissible gaps is, in the one-prime-at-a-time literature, the question of primes near large gaps in the sieved set; the classical results on primes in short intervals and the Cramér–Granville model (Granville, "Harald Cramér and the distribution of prime numbers", 1995) predict that, after conditioning on admissibility, occupancy is independent of the local arrangement of the sieve residues, which is exactly the route's null; no source measures it for twin slots at a primorial period. The permutation null used (twin labels permuted among the period's slots) is the exchangeability the "arrangement carries no occupancy information" hypothesis asserts, so a rejection would be a statement about the tile, and a null a bound on this test's power at the stated exposure. (3) Anchors: OEIS A001359 (π₂(10⁶) = 8169), D(T_19) = 378,675 and D(T_23) = 7,952,175 (Schemmel's totient at the primorial, OEIS A059861, as #1948 recorded), the per-level calibrations of #1028 (333,007 twins over 8 periods at x = 19; 1,492,887 over 2 periods at x = 23), reproduced through this instrument's code path. Exact remaining gap after this run: stated in evidence_md (the verdict at scales M/64 and M/128 and at the gap-local scale at x = 19 and 23; nothing at x = 29 or 31, whose tile words of 214,708,725 and 6,226,553,025 slots need a chunked tile sieve that this run did not build).