fix(db): drop the unused plain env_builds status index

packages/db/migrations/20260725040800_drop_env_builds_status_index.sql

@@ -0,0 +1,37 @@
+-- +goose Up
+-- +goose NO TRANSACTION
+
+-- DROP only waits on in-flight lock holders; bound the wait so it neither
+-- hangs unbounded nor dies to a short session default.
+SET statement_timeout = '1h';
+
+-- Plain btree on the raw status column: a negligible number of lifetime scans in
+-- pg_stat_user_indexes, while every build
+-- insert/supersede pays its maintenance write. Status-filtered access runs
+-- on the status_group partial indexes instead. env_builds is the single
+-- most write-amplified table (0% HOT updates, 8 indexes), so each dropped
+-- index cuts real WAL and dead-tuple volume.
+DROP INDEX CONCURRENTLY IF EXISTS public.idx_env_builds_status;
+
+-- The migrator session is reused for subsequent migrations: restore its
+-- baseline (scripts/migrator.go sets 3h per connection; RESET would clear
+-- to the unbounded server default instead).
+SET statement_timeout = '3h';
+
+-- +goose Down
+-- +goose NO TRANSACTION
+
+-- A concurrent rebuild on a ~330M row table runs for hours, and a mid-build
+-- timeout would strand an INVALID index. Rolling back is a deliberate manual
+-- operation with an operator present — unbounded on purpose.
+SET statement_timeout = 0;
+
+-- An interrupted CONCURRENTLY build leaves an INVALID index that IF NOT
+-- EXISTS would then skip forever — clear any leftover before rebuilding.
+DROP INDEX CONCURRENTLY IF EXISTS public.idx_env_builds_status;
+
+CREATE INDEX CONCURRENTLY IF NOT EXISTS idx_env_builds_status
+    ON public.env_builds (status);
+
+-- Restore the migrator session baseline (see Up).
+SET statement_timeout = '3h';