diff --git a/openfut-utas-host/tests/economy_integration.rs b/openfut-utas-host/tests/economy_integration.rs
index 11c2af1..3eec81c 100644
--- a/openfut-utas-host/tests/economy_integration.rs
+++ b/openfut-utas-host/tests/economy_integration.rs
@@ -21,13 +21,11 @@ use serde_json::{json, Value};
 /// Boot a Core instance against `db_url`, serving on an ephemeral port. Returns
 /// the serve task handle and its base URL.
 async fn start_core(db_url: &str) -> (tokio::task::JoinHandle<()>, String) {
-    // Serialize Core access with a single pooled connection so this E2E+restart
-    // harness is deterministic. NOTE: even with the WAL-establish-once +
-    // busy_timeout fixes (core 75b1830), a multi-connection pool on a brand-new
-    // DB still intermittently surfaces a write "database error" under warm-up —
-    // a deeper sqlx/SQLite pool concurrency issue that is a documented remaining
-    // blocker for the Part-T concurrency proof, not exercised by this harness.
-    let pool = openfut_core::db::init_pool(db_url, 1)
+    // Multi-connection pool: the fresh-DB write-lock race is fixed in Core
+    // (WAL-establish-once + BEGIN IMMEDIATE write transactions + busy_timeout,
+    // core fbb54ea), proven by the concurrency reproduction (800/800 concurrent
+    // writes), so a real multi-connection pool is stable here.
+    let pool = openfut_core::db::init_pool(db_url, 5)
         .await
         .expect("core pool");
     openfut_core::db::run_migrations(&pool)