Focused check of the nearby timing pair

This bounded diagnostic applies the qualified complete-cluster boundary method to one remaining stretch's specified-field-globe primary and repeat under profiled field removal. It keeps the observations, calibration bounds and scientific stopping limits unchanged.

The saved diagnosis identifies an unstable fresh correction on one side of the earlier primary's timing boundary. The new test searches the nearby boundary cluster, then checks both sides of the selected boundary. It uses two fresh smooth controls and both complete original fit endpoints; the interrupted older diagnostic remains preserved.

The finite plan permits four identities and at most 58 native evaluations, with four total numerical threads and no optimizer starts or retries. Software qualification and the exact source/input/environment freeze precede observed work. Publication and status checks perform no fits.

The current findings remain 51 globe / 0 flat / 2 unresolved. A successful diagnostic still requires a complete saved model/case/repeat review before the stretch's disposition can change.

The trial is complete: both controls passed in six native calls; both targets stopped before evaluation because of a saved-context representation error. The results retain those failures. The separately frozen corrected-context trial tests the corrected adapter; the completed allowance cannot be resumed or reused.