plan: PR-5.1.5 (c) read-side + migration 006 committed + Docker-verified; prod-apply of 006 pending user auth

Signed-off-by: "Connor Tsui" <connor@spiraldb.com>

Author: connortsui20

.big-plans/ct__bench-v4.md

@@ -12,8 +12,8 @@ current_pr: PR-5.1.5
 pr_index: 3
 outstanding_must_fix: 0
 deferred_items_total: 17
-last_user_touchpoint: 2026-06-11T19:40:00Z
-last_user_touchpoint_what: "PR-5.1.5 commit 1 (sargable) + INSTANCE UPSIZE DONE (agent ran it via bench-prod, user-authorized): db.t4g.micro->medium (4GiB), custom param group vortex-bench-pg16 (work_mem=32MB; shared_buffers ~910MB + effective_cache_size ~1.78GB auto-scaled), pgStatus in-sync, available. MEASURE-DRIVEN results (warm EXPLAIN ANALYZE on prod post-upsize): CHART PATH SOLVED — tpch chart 34.6s(micro,old) -> 13.8s(micro,sargable) -> 75ms(medium,sargable warm, all buffer hits). /api/groups REMAINING (still needed, validates comprehensive scope): (b) query-group discovery GROUP BY = 2.1s (index-only scan but HashAggregate over 4.85M; needs skip-scan/loose-index-scan); (c) DOMINANT — per-group query-summary row_number window = 6.2s for tpcds (scans 872K rows; needs index-backed latest-per-series, i.e. denormalize commit_timestamp into fact tables + supporting index, OR a recent-commits-window restriction — DESIGN FORK, touches write path); (d) collectFilterUniverse ~4s every page (DISTINCT over 4 fact tables; needs lookup/skip-scan); (e) parallelize the 64 N+1 summaries. NEXT: design migration 006 (commit_timestamp denorm for c + skip-scan indexes for b/d) + write-path populate (post-ingest.py + migrate loader) — pending user input on the (c) approach. migrations dir = repo-root migrations/."
+last_user_touchpoint: 2026-06-11T20:30:00Z
+last_user_touchpoint_what: "PR-5.1.5 PROGRESS (Docker up + user-authorized pushing): committed + Docker-verified — sargable (2e637401e), parallelize e (629b5b0b6), and (c) read-side + migration 006 + (d) indexes (680b30e6e). migration 006 denormalizes commit_timestamp into query_measurements (master-applied, requires-superuser marker) + summary index + engine/format indexes. Web vitest 211 + migrate pytest 24 green. Instance already upsized to db.t4g.medium (chart path measured 34.6s->75ms warm). PROD-APPLY of 006 is BLOCKED on explicit user auth: the harness auto-mode classifier denied 'migrate-schema.py apply' to prod RDS as a prod DB migration (status confirms 006 is the only pending; the apply is atomic + Docker-verified). REMAINING after prod-apply: (c) write-path populate (post-ingest.py + migrate/src/postgres.rs commit_timestamp), (d) collectFilterUniverse skip-scan rewrite, (b) discovery skip-scan rewrite, then push -> deploy (needs 006 applied to prod first, since the deployed summary code reads q.commit_timestamp) -> re-measure /api/groups -> gauntlet review. Master DSN: bench-prod MasterUserSecret -> PG* env + verify-full + ~/rds-ca.pem; migrate-schema.py apply."
 subagent_invocations_this_pr: 2
 subagent_invocations_total: 147
 review_cycles_this_pr: 0