Z calibration with possible upstream field removal

The first joint grid completed 27/27 fits: every tested Z-axis correction preserved globe preference, then the specified background-field preference. This separately frozen grid changes only the processing assumption, allowing some standard background field to have been removed onboard.

An IMU (inertial measurement unit) measures turning and acceleration. This uses the same complete 438-second IMU21 stretch, median speed 878.06 km/h, from ILVIS0_applanix_57300_610_LVIS_38614965.013. Z gain is fixed at −2%, zero or +2%, and constant Z offset at −1, zero or +1 degree/hour. All three models reestimate remaining parameters at each combination, with unchanged bounds, original integration, packet-header timing and the original solver. The common removal fraction starts at zero and ranges from zero to one. Each model fits it under the same limits.

The plan permits 27 starts, 200 native sweeps each, 5,400 total, no retries. The sequential launcher runs this stage then the preselected other recording. Both locks stay reserved, with one worker and two threads total. The combined ceiling is 54 starts and 10,800 sweeps, approximately 12–14 hours if both exhaust their budgets.

env PYTHONPATH=analysis OPENBLAS_NUM_THREADS=2 OMP_NUM_THREADS=2 MKL_NUM_THREADS=2 ~/venv/bin/python analysis/tests/ilvis0_z_profile_sequence.py --prepare-only

Preparation hashes sources, inputs and numerical environment; it performs no fits. The approved sequence launches with --detach instead of --prepare-only, outside a short-lived command sandbox. Sequence status and lock are in data/ilvis0-z-profile-followups-20261011/; each stage has its own manifest, authorization, start journal and completion. Do not edit either freeze while waiting or running. Interrupted charged starts remain failures; only unchanged, uncharged identities may resume. Never restart completed studies.

Results publish automatically, including failures, figures, derivatives and epoch residuals. Completed publication uses analysis/tests/ilvis0_z_profile_followup.py --study removal --package-only in the same environment; it performs no extraction or fits. Small residuals cannot identify actual onboard removal. Original field and unknown subtraction remain inseparable without further constraints. Model preferences do not determine the field's cause; statistical decisions remain abstentions.