Follow-up from #400 (question 2, ruled out of scope there). The geometry path in _intersect_mortie still calls moc_to_order once per ring — the last per-ring scalar call after the batch rewire. mortie ≥ 0.9.5 ships the ragged batch mocs_to_orders (espg/mortie#156), already consumed by the stored-index path in phase 4 of #400; adopt it on the geometry path too. Real-catalog note from #400: at order 13 the per-ring expansion is a much larger share of the remaining wall than the synthetic probe suggested (~40× bigger MOCs), so this is worth more than the earlier "roughly a third" estimate.
Follow-up from #400 (question 2, ruled out of scope there). The geometry path in
_intersect_mortiestill callsmoc_to_orderonce per ring — the last per-ring scalar call after the batch rewire. mortie ≥ 0.9.5 ships the ragged batchmocs_to_orders(espg/mortie#156), already consumed by the stored-index path in phase 4 of #400; adopt it on the geometry path too. Real-catalog note from #400: at order 13 the per-ring expansion is a much larger share of the remaining wall than the synthetic probe suggested (~40× bigger MOCs), so this is worth more than the earlier "roughly a third" estimate.