Audit (2026-05-17): 496 153 total rows, 229 261 unique by
(obj_id, poi_category, poi_name, poi_address) = ~267k duplicates (~54%).
obj_id=65136 sample: 374 rows vs 192 on site = ×2 confirmed.
Root cause: old ON CONFLICT key included snapshot_date — each scrape run
inserted fresh rows for same POI. Accumulated ×N copies = N scrape runs.
Changes:
- data/sql/110_22j_dedup_infrastructure.sql: ONE-TIME cleanup keeping min(ctid)
+ DROP old constraint + ADD UNIQUE NULLS NOT DISTINCT (PG16) for nullable
poi_category/poi_address.
- backend/app/services/scrapers/domrf_kn.py UPSERT_INFRA_SQL: ON CONFLICT
(obj_id, poi_category, poi_name, poi_address) DO NOTHING — without snapshot_date.
Closes issue #297 sub-task 22j.
