Narrow the remaining east-coordinate ranges
The saved-only comparison is complete. All 40 calculations pass their checks; the original width target passes in four cases and fails in two. Do not resume it, retry its tasks or enlarge its allowance.
- Inspect the saved east-coordinate endpoints, selected providers and remainder contributions for both specified-field cases. Separate the contribution of the older future bounds from training constraints and shared calibration. Use no new integrations or future truth to choose a favorable subset.
- Check whether the narrower direct-source future bounds already present in four of eight regions identify a useful improvement. Retain all eight regions in any final cover. A diagnostic is not an accepted partial union.
- If missing direct-source future bounds are the limiting term, investigate their runtime under a separate bounded profile. The previously proposed omission of unused endpoint force certificates is a candidate, not an established speedup. Preserve endpoint rotation state and qualify exact equivalence before using any faster implementation.
- Freeze any further certificate generation and complete-cover comparison separately, with finite costs and the original target. No extension of a completed study follows from this plan.
- After the short-example target passes, test useful ranges at recording length and assess floating execution and observation-error assumptions. Those are still part of V01; new withheld prediction fits remain blocked.