Longer inputs and clock shifts reach the calculation correctly
All six interface cases pass in 109 seconds. The generated inputs now span eight minutes, long enough to cover the earlier 438-second prediction plan. Their original four-minute measurement prefix is preserved exactly.
The new software-test clock bound is ±1 second. The old and current callers agree exactly on positions, all 27 calibration-derivative columns and warning flags at the declared long-duration queries. The checked calls return no warnings, and the clock derivative remains present. All other calibration limits and initial settings are preserved.
Both callers use the same numerical kernel. This verifies input handling and clock support, not the physical model independently. It also does not supply useful nonlinear prediction ranges or a validated measurement-error law. The ±1-second clock bound is a declared qualification assumption, not a measured instrument specification.
The next task is to constrain long predictions using their own earlier observations, retaining every constraint through later refinements. No aircraft fit or protected ESA observation enters these controls.