Reproducibility question: canonical VIIRS source and discovery-time contract

[tlogandavis-crypto]

Hello WildfireIA maintainers, I am independently auditing the discovery-time input contract of the publicly released WildfireIA benchmark. This is a reproducibility request, not an assertion that the benchmark contains an error.

Could you please clarify:

1. Which exact code commit or script revision constructed the released canonical VIIRS features? Was the public pipeline.py implementation the version actually used for the release?
2. Which original NASA FIRMS sources, sensors, processing versions, and source snapshots were used? If available, could you provide a version identifier, checksum, or construction manifest?
3. What timezone convention applies to the underlying FPA-FOD DISCOVERY_DATE and DISCOVERY_TIME fields? How were missing discovery times, apparent 00:00 values, UTC conversion, and daylight-saving time handled?
4. Were VIIRS detection acquisition times filtered against each individual fire's discovery timestamp, or were detections matched by calendar date and spatial proximity? Were original publication/ingest timestamps retained?
5. For FOD_ID 400645736, the released canonical feature table contains two VIIRS detections within 1 km. Our retrospective lookup using current standard-processed SNPP and NOAA-20 sources returns four candidate matches. Could you identify the original two detections (sensor, acquisition timestamp, coordinates, FRP, or stable detection/granule ID) and any additional filtering or nearest-fire assignment rule? The audit uses a deliberately selected 30-event sample and does not establish a population leakage rate. Original source identity and timestamp semantics would help us distinguish genuine temporal issues from source-version differences. Thank you for your work and for any provenance details you can share.