diff --git a/zebra/connected.c b/zebra/connected.c
index 3b9ebe14a4..8c4ba163bd 100644
--- a/zebra/connected.c
+++ b/zebra/connected.c
@@ -393,10 +393,10 @@ void connected_down(struct interface *ifp, struct connected *ifc)
 	 * head.
 	 */
 	rib_delete(afi, SAFI_UNICAST, zvrf->vrf->vrf_id, ZEBRA_ROUTE_CONNECT, 0,
-		   0, &p, NULL, &nh, 0, zvrf->table_id, 0, 0, false);
+		   0, &p, NULL, &nh, 0, zvrf->table_id, 0, 0, false, true);
 
 	rib_delete(afi, SAFI_MULTICAST, zvrf->vrf->vrf_id, ZEBRA_ROUTE_CONNECT,
-		   0, 0, &p, NULL, &nh, 0, zvrf->table_id, 0, 0, false);
+		   0, 0, &p, NULL, &nh, 0, zvrf->table_id, 0, 0, false, true);
 
 	/* Schedule LSP forwarding entries for processing, if appropriate. */
 	if (zvrf->vrf->vrf_id == VRF_DEFAULT) {
diff --git a/zebra/kernel_socket.c b/zebra/kernel_socket.c
index 4c29b999f0..02963651a0 100644
--- a/zebra/kernel_socket.c
+++ b/zebra/kernel_socket.c
@@ -1136,7 +1136,7 @@ void rtm_read(struct rt_msghdr *rtm)
 	if (rtm->rtm_type == RTM_CHANGE)
 		rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL,
 			   0, zebra_flags, &p, NULL, NULL, 0, RT_TABLE_MAIN, 0,
-			   0, true);
+			   0, true, false);
 	if (rtm->rtm_type == RTM_GET || rtm->rtm_type == RTM_ADD
 	    || rtm->rtm_type == RTM_CHANGE)
 		rib_add(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL, 0,
@@ -1145,7 +1145,7 @@ void rtm_read(struct rt_msghdr *rtm)
 	else
 		rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL,
 			   0, zebra_flags, &p, NULL, &nh, 0, RT_TABLE_MAIN, 0,
-			   0, true);
+			   0, true, false);
 }
 
 /* Interface function for the kernel routing table updates.  Support
diff --git a/zebra/redistribute.c b/zebra/redistribute.c
index 0b97562424..c0f89e6afe 100644
--- a/zebra/redistribute.c
+++ b/zebra/redistribute.c
@@ -727,7 +727,7 @@ int zebra_del_import_table_entry(struct zebra_vrf *zvrf, struct route_node *rn,
 	rib_delete(afi, SAFI_UNICAST, zvrf->vrf->vrf_id, ZEBRA_ROUTE_TABLE,
 		   re->table, re->flags, &p, NULL, re->nhe->nhg.nexthop,
 		   re->nhe_id, zvrf->table_id, re->metric, re->distance,
-		   false);
+		   false, false);
 
 	return 0;
 }
diff --git a/zebra/rib.h b/zebra/rib.h
index b9f4e56905..be680a112f 100644
--- a/zebra/rib.h
+++ b/zebra/rib.h
@@ -379,7 +379,7 @@ extern void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type,
 		       unsigned short instance, int flags, struct prefix *p,
 		       struct prefix_ipv6 *src_p, const struct nexthop *nh,
 		       uint32_t nhe_id, uint32_t table_id, uint32_t metric,
-		       uint8_t distance, bool fromkernel);
+		       uint8_t distance, bool fromkernel, bool connected_down);
 
 extern struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id,
 				     union g_addr *addr,
diff --git a/zebra/rt_netlink.c b/zebra/rt_netlink.c
index 4a6839d3b1..50b1a62d86 100644
--- a/zebra/rt_netlink.c
+++ b/zebra/rt_netlink.c
@@ -837,7 +837,7 @@ static int netlink_route_change_read_unicast(struct nlmsghdr *h, ns_id_t ns_id,
 		if (nhe_id) {
 			rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0, flags,
 				   &p, &src_p, NULL, nhe_id, table, metric,
-				   distance, true);
+				   distance, true, false);
 		} else {
 			if (!tb[RTA_MULTIPATH]) {
 				struct nexthop nh;
@@ -847,13 +847,13 @@ static int netlink_route_change_read_unicast(struct nlmsghdr *h, ns_id_t ns_id,
 					gate, afi, vrf_id);
 				rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0,
 					   flags, &p, &src_p, &nh, 0, table,
-					   metric, distance, true);
+					   metric, distance, true, false);
 			} else {
 				/* XXX: need to compare the entire list of
 				 * nexthops here for NLM_F_APPEND stupidity */
 				rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0,
 					   flags, &p, &src_p, NULL, 0, table,
-					   metric, distance, true);
+					   metric, distance, true, false);
 			}
 		}
 	}
diff --git a/zebra/zapi_msg.c b/zebra/zapi_msg.c
index ec61be0b39..4b31b46bce 100644
--- a/zebra/zapi_msg.c
+++ b/zebra/zapi_msg.c
@@ -1890,7 +1890,7 @@ static void zread_route_del(ZAPI_HANDLER_ARGS)
 
 	rib_delete(afi, api.safi, zvrf_id(zvrf), api.type, api.instance,
 		   api.flags, &api.prefix, src_p, NULL, 0, table_id, api.metric,
-		   api.distance, false);
+		   api.distance, false, false);
 
 	/* Stats */
 	switch (api.prefix.family) {
diff --git a/zebra/zebra_rib.c b/zebra/zebra_rib.c
index 5fcb513e3c..ff30de18a3 100644
--- a/zebra/zebra_rib.c
+++ b/zebra/zebra_rib.c
@@ -3041,7 +3041,7 @@ void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type,
 		unsigned short instance, int flags, struct prefix *p,
 		struct prefix_ipv6 *src_p, const struct nexthop *nh,
 		uint32_t nhe_id, uint32_t table_id, uint32_t metric,
-		uint8_t distance, bool fromkernel)
+		uint8_t distance, bool fromkernel, bool connected_down)
 {
 	struct route_table *table;
 	struct route_node *rn;
@@ -3163,7 +3163,8 @@ void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type,
 					    rn, fib,
 					    zebra_route_string(fib->type));
 			}
-			if (allow_delete) {
+			if (allow_delete
+			    || CHECK_FLAG(dest->flags, RIB_ROUTE_ANY_QUEUED)) {
 				UNSET_FLAG(fib->status, ROUTE_ENTRY_INSTALLED);
 				/* Unset flags. */
 				for (rtnh = fib->nhe->nhg.nexthop; rtnh;
@@ -3248,6 +3249,19 @@ void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type,
 		rib_delnode(rn, same);
 	}
 
+	/*
+	 * This is to force an immediate re-eval of this particular
+	 * node via nexthop tracking.  Why?  Because there are scenarios
+	 * where the interface is flapping and the normal queuing methodology
+	 * will cause down/up events to very very rarely be combined into
+	 * a non-event from nexthop tracking perspective.  Leading
+	 * to some fun timing situations with upper level routing protocol
+	 * trying to and failing to install routes during this blip.  Especially
+	 * when zebra is under load.
+	 */
+	if (connected_down)
+		zebra_rib_evaluate_rn_nexthops(rn,
+					       zebra_router_get_next_sequence());
 	route_unlock_node(rn);
 	return;
 }