Changed
- Parquet sync now skips noncompliant datasets (instead of failing full runs) when source artifacts are missing or unusable.
- Sync outputs now include
skipped_detailswith per-dataset reason and skip type (resumeornoncompliant).
Added
- Test coverage for noncompliant skip behavior in parquet sync.
Why this release
This keeps large full-sync workflows stable while still surfacing exactly which upstream datasets are noncompliant.
See CHANGELOG.md for full details.