Constant part of the substituted position row, millimetres.
Common independent-control lower bounds, millimetres.
Positive safe cap on supplied terms, including repeated and canceled addresses.
One private positional observation, millimetres; it need not be inside the box.
Original independent IDs and dimensionless weights; repeated IDs combine in input order.
Positive millimetre tolerance; exact equality is not infeasible.
Common independent-control upper bounds, millimetres.
Binary64 sum of absolute combined weights, in first-address order.
Predicate scope and diagnostic precision qualification.
Exact strict distance-squared comparison for the combined binary64 row.
Outward-rounded lower endpoints, millimetres.
Approximate distance to the exact row image, millimetres.
Unique original addresses, including exact-zero cancellations.
Outward-rounded upper endpoints, millimetres.
Binary64 sum of combined weights, in first-address order.
Bound one declared affine row over a common independent-coefficient box. Duplicate original IDs first combine in input order using binary64 addition. The resulting floating-point coefficients, constants, bounds and target are then interpreted as exact dyadic numbers. Endpoint arithmetic and the strict squared-distance comparison against tolerance are exact for that row.
Returned endpoints round outward. minimumDistanceMm is an approximate number view; only infeasibleAtTolerance uses the exact comparison. A false decision does not establish that several observations can be fitted simultaneously. Row identities do not identify source vertices, and this box is not generated geometry, continuous source support or a closest-surface distance certificate. Nonfinite input, conflicting bounds, invalid IDs, exhausted raw term budgets, nonfinite duplicate sums or nonrepresentable number views throw atomically.