Corrected receiver context for the nearby timing pair
The first focused trial passed both controls but stopped both targets before native evaluation because their attached receiver context used dictionaries where the worker expected numerical latitude/longitude/height rows. Its failed starts and six native calls remain in the original record.
The corrected adapter checks every receiver coordinate and relative timestamp against the frozen selection and first IMU interval start. It preserves the original residuals, Jacobians and fitted parameters. Tests cover the complete 554-epoch context for both target endpoints before any original-file extraction.
The separate finite trial retains the same observations, models, bounds and full stopping checks: two fresh controls and the missing primary/repeat pair, at most 58 native calls, zero optimizer starts, four total numerical threads and no retries. A complete result still needs a saved model/case/repeat review. Current findings remain 51 globe / 0 flat / 2 unresolved.
The trial is complete: 54 native calls, both controls passed, no software errors. Both targets agree closely on repeat but fail the fresh right-side calibration correction. The saved linear-system diagnosis reproduces the failure before nonlinear propagation. The completed allowance cannot be resumed or reused.