perf(db): drop redundant clinical_registry_record_sources_record_idx

supabase/drift-manifest.json

@@ -1,9 +1,9 @@
 {
-  "generated_at": "2026-07-09T10:38:15.859Z",
+  "generated_at": "2026-07-11T02:45:10.714Z",
   "generator": "scripts/generate-drift-manifest.ts",
   "postgres_image": "supabase/postgres:17.6.1.127",
-  "schema_sha256": "3a60bf1561142aed58e45a1cf8a9605ccc715dd7358e154c4ad9879b82be4cd5",
-  "replay_seconds": 20,
+  "schema_sha256": "aa8a419ab55b4637b858df628aa250b714ab1acd715d4f7786bcf6ec71f5e218",
+  "replay_seconds": 22,
   "snapshot": {
     "views": [
       {
@@ -4263,12 +4263,6 @@
         "table": "clinical_registry_record_sources",
         "def_hash": "aff03b1509bdec54ffc6611caf47b647"
       },
-      {
-        "def": "CREATE INDEX clinical_registry_record_sources_record_idx ON public.clinical_registry_record_sources USING btree (record_id)",
-        "name": "clinical_registry_record_sources_record_idx",
-        "table": "clinical_registry_record_sources",
-        "def_hash": "ef53891aac4bf949c894e35187d71517"
-      },
       {
         "def": "CREATE UNIQUE INDEX clinical_registry_records_owner_id_kind_slug_key ON public.clinical_registry_records USING btree (owner_id, kind, slug)",
         "name": "clinical_registry_records_owner_id_kind_slug_key",

supabase/migrations/20260711000000_drop_redundant_registry_sources_record_index.sql

@@ -0,0 +1,19 @@
+-- Drop redundant index (supabase-postgres-best-practices review, 2026-07-11).
+--
+-- clinical_registry_record_sources_record_idx (record_id) is a strict prefix of
+-- the index backing the UNIQUE (record_id, document_id) constraint on
+-- clinical_registry_record_sources. PostgreSQL can use the unique index's
+-- leading column for every equality lookup on record_id, so the standalone
+-- index adds write overhead with no read benefit. Same rationale as
+-- 20260702110000_drop_redundant_indexes.sql.
+--
+-- clinical_registry_record_sources_document_idx stays: document_id is not a
+-- leading column of any other index and backs the documents ON DELETE CASCADE.
+--
+-- NOTE: DROP INDEX CONCURRENTLY cannot run inside a transaction block, and
+-- Supabase migrations run in one. The table is small (curated registry links),
+-- so a plain DROP INDEX is fine. For zero-impact removal run manually outside
+-- a transaction:
+--   DROP INDEX CONCURRENTLY IF EXISTS public.clinical_registry_record_sources_record_idx;
+
+drop index if exists public.clinical_registry_record_sources_record_idx;

supabase/schema.sql

@@ -4945,8 +4945,9 @@ create table if not exists public.clinical_registry_record_sources (
 
 create index if not exists clinical_registry_records_owner_kind_idx
   on public.clinical_registry_records(owner_id, kind, title);
-create index if not exists clinical_registry_record_sources_record_idx
-  on public.clinical_registry_record_sources(record_id);
+-- No standalone record_id index: the unique (record_id, document_id) index's
+-- leading column serves record_id lookups (20260711000000 dropped the redundant
+-- clinical_registry_record_sources_record_idx).
 create index if not exists clinical_registry_record_sources_document_idx
   on public.clinical_registry_record_sources(document_id);