Original audit · Approved plan
DAG 345 validation job 2377.0 (skycell.2487.034) exceeded its 16384 MB cgroup limit. This CLI continuation cannot access Condor networking; it made no job changes. At 01:14:11 UTC on October 4, saved metadata reports 975/975 exact cell replays and 815/975 recipient validations; 160 recipient reports and full-template reduction remain outstanding. These counts are a metadata audit, not revalidated payload checksums or catalogue completeness.
Diagnosis, evidence, and scoped recovery instructions · Saved audit. Existing pilot figures do not certify the full field. Conclusion: incomplete; scheduler access and a single-task memory trial are needed.
Live C4 backfill and validation progress
codex/ps1-removal-ledger; campaign pin d3549ee pushed. No production merge or template replacement. Full-field backfill and all-seam certification remain unfinished.The recorder captures actual deletion support and separates finite flux changes, already-zero pixels, and invalid pixels. The pilot reuses saved segmentation where available. PS1 DR2 stack catalogues retain separate measurements and object IDs, including non-primary detections. Gaia has no photometric cut. Identity matching preserves ambiguous cases and missing epoch/motion information.
| Field | Cell | Exact replay | Saved segmentation reused | Catalogue measurements (not unique stars) | Evidence |
|---|---|---|---|---|---|
| C4 | skycell.2390.080 | True | False | 33,573 | record |
| F1 | skycell.2484.041 | True | True | 18,820 | record |
| C4 | skycell.2528.005 | True | True | 23,075 | record |
| F2 | skycell.2611.018 | True | True | 18,100 | record |
For C4 2528.005 all 21 previously omitted Gaia IDs are now present and associated with explicit deletion. Source table · Region associations.
Examples at PS1 detection positions without a confirmed Gaia identity. Some may be blends or artifacts near saturated stars; catalogue rows are not automatically certified as additional physical stars. These panels show raw, pre-deletion, post-deletion and deleted images on a shared scale.
Restoring the measured explicit deletions in C4 cell 2528.005, with all neighbouring combined cells fixed, passes through the historical v2 ownership and the actual cross-projection full-halo correction. Reconstructed published pixels are exactly equal; the transported deletion agrees with the paired render within the declared float32 arithmetic bound. Complete operator check.
This verifies one scoped intervention, not the entire CCD. Other recipients consuming this donor and all other seam cases still need coverage. A known pre-existing F2 cross-projection image disagreement remains explicitly expected to fail in the inherited tests; it is not claimed fixed.
Cache-based centre-cell pilots across all five paper fields passed exact image replay. The full 975-cell C4 backfill and 975 recipient validations are running as independent DAG 345 on plscience3.stsci.edu, followed automatically by full-template reduction. Catalogue-to-image calibration and lossless independent catalogue-count checks are implemented. Full donor/recipient coverage is now being processed. Production integration and full-field scientific review remain pending. Candidate support circles do not yet establish the fraction of each stellar PSF removed.
Run record and commands · Product/cache inventory · Isolated code branch
Conclusion: the new cell ledger and a real historical seam transport have passed their first tests; the repair is in progress and is not yet a validated replacement for the full paper ledger.