{"state": "statistics_valid_report_scope_issues_to_be_fixed_in_new_v2", "issues": ["A target flagged only in all1559 but included in294 can have its flags hidden by training-scope report choice, yielding an empty case reason", "Some near-zero-own-distance medoid relative ratios become huge negative values; no outlier flag uses them as positive evidence", "A medoid example with only one distinct acquisition-group reference can compare to itself"], "labels_or_dataset_changed": false, "old_artifacts_preserved": true}