Aggregation: cross-run story dedup
Stop the same ongoing news story from accumulating multiple separate aggregates. Root causes found in prod DB analysis (Oct 2026): the incremental run only sees the 100 most-recently-updated aggregates (~9 h of coverage; 111 near-duplicate aggregate-name pairs in the DB), and the consolidation pass never merges duplicate candidate topics. The two issues here are one combined fix: relevance pre-filter for candidate visibility + merge-aware consolidation.
No due date
100% Completed
