diff --git a/packages/db/migrations/007_techniques_in_uniqueness.sql b/packages/db/migrations/007_techniques_in_uniqueness.sql
new file mode 100644
index 000000000..c6e8a5441
--- /dev/null
+++ b/packages/db/migrations/007_techniques_in_uniqueness.sql
@@ -0,0 +1,32 @@
+-- ============================================================
+-- Migration 007 — include techniques in result-row uniqueness
+-- ============================================================
+--
+-- Migration 006 made techniques a per-measurement dimension but left
+-- benchmark_results_unique and eval_results_unique unchanged. Result: two
+-- measurements at the same (workflow_run, config, isl, osl, conc) but with
+-- DIFFERENT techniques (e.g. mtp_layers=4 vs 6 — same gemma4 deployment, two
+-- MTP variants) collide on insert and the later one silently overwrites the
+-- earlier one.
+--
+-- Hot-fix discovered during the first post-006 ingest: gemma4 run 26387431571
+-- landed only 6 of ~24 expected rows because n4 and n6 variants now share a
+-- config_id and clobbered each other.
+--
+-- Fix: add the techniques jsonb to both unique constraints. Postgres b-tree
+-- works on jsonb (used in latest_benchmarks PK already), so this is a clean
+-- add. Also re-truncate so the re-ingest after this migration is from scratch.
+
+alter table benchmark_results drop constraint benchmark_results_unique;
+alter table benchmark_results add constraint benchmark_results_unique
+  unique (workflow_run_id, config_id, benchmark_type, isl, osl, conc, techniques);
+
+alter table eval_results drop constraint eval_results_unique;
+alter table eval_results add constraint eval_results_unique
+  unique (workflow_run_id, config_id, task, isl, osl, conc, techniques);
+
+-- Wipe partially-clobbered data; auto-ingest will repopulate.
+truncate benchmark_results, eval_results, eval_samples, run_stats,
+         availability, changelog_entries, server_logs, workflow_runs
+  restart identity cascade;
+truncate configs restart identity cascade;
diff --git a/packages/db/src/etl/benchmark-ingest.ts b/packages/db/src/etl/benchmark-ingest.ts
index f61fc1a8e..258138f09 100644
--- a/packages/db/src/etl/benchmark-ingest.ts
+++ b/packages/db/src/etl/benchmark-ingest.ts
@@ -29,9 +29,16 @@ export async function bulkIngestBenchmarkRows(
 
   // Postgres rejects ON CONFLICT DO UPDATE if the same conflict key appears
   // more than once in a single batch. Deduplicate within the batch, keeping
-  // the last occurrence (last metrics for each unique config/isl/osl/conc).
+  // the last occurrence. The conflict key must include techniques — two MTP
+  // variants at the same (config, isl, osl, conc) are legitimately distinct
+  // measurements and must not collapse to one row (migration 007).
   const seen = new Map<string, BenchmarkParams & { configId: number }>();
-  for (const r of rows) seen.set(`${r.configId}-${r.isl}-${r.osl}-${r.conc}`, r);
+  for (const r of rows) {
+    // parseTechniques() builds the techniques object with deterministic key
+    // order, so a plain JSON.stringify is a stable dedup discriminator.
+    const techniquesKey = JSON.stringify(r.techniques);
+    seen.set(`${r.configId}-${r.isl}-${r.osl}-${r.conc}-${techniquesKey}`, r);
+  }
   const deduped = [...seen.values()];
 
   const configIds = deduped.map((r) => r.configId);
@@ -58,11 +65,10 @@ export async function bulkIngestBenchmarkRows(
       unnest(${sql.array(images)}),
       unnest(${sql.array(metricsJsons)}::jsonb[]),
       unnest(${sql.array(techniquesJsons)}::jsonb[])
-    on conflict (workflow_run_id, config_id, benchmark_type, isl, osl, conc)
+    on conflict (workflow_run_id, config_id, benchmark_type, isl, osl, conc, techniques)
     do update set
       metrics = excluded.metrics,
-      image = excluded.image,
-      techniques = excluded.techniques
+      image = excluded.image
     returning (xmax = 0) as inserted, id
   `;
 
diff --git a/packages/db/src/etl/eval-ingest.ts b/packages/db/src/etl/eval-ingest.ts
index c01715014..b607319b2 100644
--- a/packages/db/src/etl/eval-ingest.ts
+++ b/packages/db/src/etl/eval-ingest.ts
@@ -38,10 +38,9 @@ export async function ingestEvalRow(
       ${p.isl}, ${p.osl}, ${p.conc}, ${p.lmEvalVersion},
       ${sql.json(p.metrics)}, ${sql.json(p.techniques)}
     )
-    on conflict (workflow_run_id, config_id, task, isl, osl, conc)
+    on conflict (workflow_run_id, config_id, task, isl, osl, conc, techniques)
     do update set
-      metrics = excluded.metrics,
-      techniques = excluded.techniques
+      metrics = excluded.metrics
     returning id, (xmax = 0) as inserted
   `;
   return { outcome: row.inserted ? 'new' : 'dup', id: row.id };