Generate only the twelve missing future certificates
Prepare and launch only after the separate six-case serial execution screen passes with exact report agreement. Use the unchanged direct-source mathematics under that one-worker, one-thread recipe. Do not adopt the failed endpoint shortcut or restart the earlier complete-comparison study.
Select all twelve identities whose original direct-source task ended solely because of its time limit and has no saved seventh (future) bound. Determine this from sealed task metadata before calculation, without model costs or future truth. Preserve all original 1,600-packet inputs, 7.2-second queries, orientation sectors, sources, references and child-domain proofs.
Only the future certificate is new. Reuse the already completed training certificates later; do not recalculate them here. Where a compatible older future certificate exists, retain it and check the new maps and nonwidening remainders against it. Missing older reports remain explicitly unavailable.
New finite allowance
- Twelve identities and twelve bound calls; 172,800 logical bound steps.
- Twelve reference setups and child proofs; 72 parent support checks.
- Twelve source/derivative/point setups and at most 19,200 visits in each separately reserved stream, with the original per-bound source recipe.
- One worker, one numerical thread; no other numerical study active.
- Preserve the 120-second cooperative per-bound ceiling. This new serial study has a 1,500-second overall ceiling. No retries, signals or expansion.
- Zero native trajectory oracles, LP proposals, optimizer starts, original aircraft reads or protected ESA outcomes.
All twelve certificates must complete and pass their checks. This is a new execution freeze for missing queries, not a relabeling of earlier timeouts. The sector-generalized adapter must otherwise retain the qualified preflight calculation unchanged. Freeze its sources, selection, exact provider hashes and environment before launch.
After successful generation, a separately frozen saved-only comparison must rebuild and audit complete region covers. Individual certificates cannot by themselves pass the full width target. Keep all earlier failure labels, the original width target and the unqualified observational-coverage status.