Eight-minute input and clock handling

6/6 API cases pass; 0 unstarted. Elapsed 108.63 seconds.

Each generated history contains 96,000 packets, retaining the original four-minute prefix bit for bit. The new clock-offset bound is ±1 second; all other physical settings are preserved.

Charged 24/24 native queries, no interval bounds or optimizer starts. Queries at 438 and 478 seconds retain the complete declared clock support. Passing cases match positions, all 27 Jacobian columns and warning bits exactly, with nonzero clock derivatives.

Both callers use the same compiled kernel. This is input/API qualification, not independent physical validation, useful nonlinear prediction ranges or an established observation-error law. No aircraft or protected ESA observations enter the study.

No new numerical work is performed by this saved-only report builder. It preserves the sealed outcomes and charges. Rebuild with build_report.py; build_report.py --check verifies publication.

Frozen scope · Full-horizon requirements.