diff --git a/orte/mca/rml/base/rml_base_contact.c b/orte/mca/rml/base/rml_base_contact.c
index 73ed100473..fa87289882 100644
--- a/orte/mca/rml/base/rml_base_contact.c
+++ b/orte/mca/rml/base/rml_base_contact.c
@@ -68,6 +68,7 @@ int orte_rml_base_update_contact_info(opal_buffer_t* data)
     char *rml_uri;
     orte_process_name_t name;
     int rc;
+    orte_jobid_t jobid;
 
     /* unpack the data for each entry */
     num_procs = 0;
@@ -96,6 +97,12 @@ int orte_rml_base_update_contact_info(opal_buffer_t* data)
          * since we were given the contact info
          */
         orte_routed.update_route(&name, &name);
+        /* we only get an update from a single jobid - the command
+         * that creates these doesn't cross jobid boundaries - so
+         * record it here
+         */
+        jobid = name.jobid;
+        /* track how many procs were in the message */
         ++num_procs;
     }
     if (ORTE_ERR_UNPACK_READ_PAST_END_OF_BUFFER != rc) {
@@ -103,12 +110,13 @@ int orte_rml_base_update_contact_info(opal_buffer_t* data)
         return rc;
     }    
     
-    /* if we are a daemon, this update would include updated contact info
+    /* if we are a daemon and this was info about our jobid, this update would
+     * include updated contact info
      * for all daemons in the system - indicating that the number of daemons
      * changed since we were initially launched. Thus, update the num_procs
      * in our process_info struct so we can correctly route any messages
      */
-    if (orte_process_info.daemon) {
+    if (ORTE_PROC_MY_NAME->jobid == jobid && orte_process_info.daemon) {
         orte_process_info.num_procs = num_procs;
     }
     
diff --git a/orte/mca/routed/unity/routed_unity.c b/orte/mca/routed/unity/routed_unity.c
index cf224cd4d1..91ff6b32c0 100644
--- a/orte/mca/routed/unity/routed_unity.c
+++ b/orte/mca/routed/unity/routed_unity.c
@@ -205,7 +205,6 @@ static int process_callback(orte_jobid_t job, opal_buffer_t *buffer)
     orte_std_cntr_t cnt;
     char *rml_uri;
     int rc;
-    orte_rml_cmd_flag_t command=ORTE_RML_UPDATE_CMD;
     
     /* lookup the job object */
     if (NULL == (jdata = orte_get_job_data_object(job))) {
@@ -266,30 +265,6 @@ static int process_callback(orte_jobid_t job, opal_buffer_t *buffer)
             jdata->state = ORTE_JOB_STATE_RUNNING;
         }
         
-        /* first update the daemons so they will know how to talk to the
-         * procs - this is required for support of modex and barrier
-         */
-        OBJ_CONSTRUCT(&buf, opal_buffer_t);
-        /* pack an update command */
-        if (ORTE_SUCCESS != (rc = opal_dss.pack(&buf, &command, 1, ORTE_RML_CMD))) {
-            ORTE_ERROR_LOG(rc);
-            OBJ_DESTRUCT(&buf);
-            return rc;
-        }
-        /* pack the RML contact info for each proc */
-        if (ORTE_SUCCESS != (rc = orte_rml_base_get_contact_info(jdata->jobid, &buf))) {
-            ORTE_ERROR_LOG(rc);
-            OBJ_DESTRUCT(&buf);
-            return rc;
-        }
-        /* send it to the daemons via xcast */
-        if (ORTE_SUCCESS != (rc = orte_grpcomm.xcast(ORTE_PROC_MY_NAME->jobid, &buf, ORTE_RML_TAG_RML_INFO_UPDATE))) {
-            ORTE_ERROR_LOG(rc);
-            OBJ_DESTRUCT(&buf);
-            return rc;
-        }
-        OBJ_DESTRUCT(&buf);
-
         /* now send to the procs so they release from their barrier */
         OBJ_CONSTRUCT(&buf, opal_buffer_t);
         /* pack the RML contact info for each proc */