Guidance-track orientation under geometric uncertainty: modest coverage gains without demonstrated benefit from confirmation

Changing straight guidance-track orientation can reduce geometric coverage deviation, but whether a separate confirmation stage improves orientation decisions has remained unclear. Under an externally timestamped protocol, we evaluated 360 synthetic fields and 350 real field boundaries with simulated execution, reserved headlands, prespecified operational budgets and separate calibration, confirmation, reference-search and holdout scenarios. Relative to the nominal longest-edge direction, the full rule reduced mean summed geometric deviation by 0.102–0.120 percentage points (10.2–12.0 m²/ha) across the four primary cohort/error settings. However, confirmation did not improve on calibration-only selection: under the low-position-error model, both rules chose identical angles on all 710 fields; under moderate error, they differed on only 3/360 synthetic fields and 5/350 real boundaries, with slightly unfavorable full-rule point estimates and paired intervals reaching or crossing zero. No point-adverse primary switches were observed for the full rule, calibration-only rule or budgeted deterministic comparator, so prevention of harmful switches was not demonstrated. Weighted-loss, angle-grid, swath-phase and budget sensitivities reveal that results depend on geometric and operational assumptions. A retrospective polygon-overlay correction invalidated an earlier harmful-comparator example. The findings support modest, model-conditional benefits of reorientation, but no incremental value from confirmation or measured machinery performance.