diff --git a/src/controller/cluster.c b/src/controller/cluster.c
index d24bb28..6799ff1 100644
--- a/src/controller/cluster.c
+++ b/src/controller/cluster.c
@@ -1081,8 +1081,13 @@ int cluster_start(cluster_t* c)
         return rc;
     }
 
-    iomgr_start(&(c->iomgr));
-    vtaskmgr_start(&(c->vtaskmgr));
+    /* Guard: myapp_run() starts iomgr and vtaskmgr before election so that
+     * incoming OP_requestVote RPCs are accepted during the election phase.
+     * Only start here if they were not already started (legacy cluster_init path). */
+    if (c->iomgr.worker == 0)
+        iomgr_start(&(c->iomgr));
+    if (c->vtaskmgr.dispatcher == 0)
+        vtaskmgr_start(&(c->vtaskmgr));
     delaymgr_start(&(c->delaymgr));
 
     /* Start persistence fsync thread and register WAL hooks */
diff --git a/src/controller/myapp.c b/src/controller/myapp.c
index 9fcd0d7..837082b 100644
--- a/src/controller/myapp.c
+++ b/src/controller/myapp.c
@@ -68,6 +68,13 @@ int myapp_run(myconf_t* myconf)
         vtaskmgr_init(&(c.vtaskmgr), nvcpus);
     }
 
+    /* Req 1.2: start the I/O worker thread so that incoming OP_requestVote RPCs
+     * from peers are accepted and dispatched during the election phase.
+     * Without iomgr_start here, connections queue in the kernel (Recv-Q > 0)
+     * but no thread reads them, causing peers to time out waiting for vote replies. */
+    iomgr_start(&(c.iomgr));
+    vtaskmgr_start(&(c.vtaskmgr));
+
     /* Req 3.1–3.4: probe peers for config tags before election so that a node
      * joining an existing cluster knows to defer its config fetch to post-election
      * rather than participating as a first-boot candidate with no config. */