Saved timing-boundary review
Question
Do the two failed fits in the fresh calibration place receiver queries close to internal IMU integration boundaries? Proximity would guide a later diagnostic; it would not establish the cause of the failure or qualify either fit.
Fixed scope
Read only the two saved starts for seed 281474977010693, rate-sample errors, specified-field globe, with field removal allowed. Use the closed calibration manifest, summary, artifact hashes and unchanged failure labels.
The shared lll.prediction_phase_review.review_clock routine reconstructs only
the declared uniform clock. It must reproduce the saved little-endian float64
hashes of every interval duration and query time before reporting distances.
It generates no inertial observations, receiver positions, truth or noise.
One saved-data review, at most two fitted records, one thread and a ten-minute cooperative ceiling. No optimizer starts, native calls, noise draws, retries, threshold changes or use of protected aircraft data. Known-answer tests check boundary hits, both signs, invalid clocks, hash mismatches and retained failures.
Deliverables and next gate
Preserve a source-bound report and per-query CSV. Report the numerical stopping failure separately from timing proximity. Any native derivative probe, revised solver or further calibration needs its own reviewed finite plan.