Online Resource 1 for "Learning to Mask under Drift: Online Adaptation of Dual-Mask Reconstruction for Multivariate Time Series Anomaly Detection" — code, checkpoints, trajectories, and raw results

v1,Supplementary material (Online Resource 1) for the revised manuscript "Learning to Mask under Drift" (submitted to Applied Intelligence).Contains: (A) all experiment code and reproduction scripts, including the affiliation/VUS metric packages; (B) per-run metric JSON files for 1,518 runs (28 SMD entities, 54 SMAP and 26 MSL channels, three seeds); (C1–C2) per-point score trajectories for the same runs (source ofFigures 1–4); (D01–D15) the frozen pre-drift checkpoints from which every online run starts; (E) the complete pilot study results and logs. Every number in the paper's tables can be recomputed from these files with the included aggregation scripts; see README.md for the layout and instructions. Benchmarks (SMD/SMAP/MSL) are public and not redistributed. v2 adds OnlineResource1_F_daca_replay.zip` — companion volume for the DACA follow-up study (alarm-budget control): causal replay harness, all threshold policies, the anomaly-dense stress-test generator, the raw per-trajectory result files underlying every table and figure of that paper, the extended-sensitivity runner (sens_ext.py, Page–Hinkley delta/lambda and replay cadence), and a 336-check numeric audit script (audit_paper.py) with its full report (results/audit_report.md). Requires only the C1/C2 trajectory zips from this record; no model training needed. v3 adds the ELSA companion artifacts (eventizer design study): 18 result files, 13 reproduction scripts, numbers.json, README with regeneration order.