diff --git a/doc/bird.sgml b/doc/bird.sgml
index b12ac544..2bce3d57 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -675,21 +675,6 @@ to set options.
 	disadvantage is that trie-enabled routing tables require more memory,
 	which may be an issue especially in multi-table setups. Default: off.
 
-	<tag><label id="rtable-min-settle-time">min settle time <m/time/</tag>
-	Specify a minimum value of the settle time. When a ROA table changes,
-	automatic <ref id="proto-rpki-reload" name="RPKI reload"> may be
-	triggered, after a short settle time. Minimum settle time is a delay
-	from the last ROA table change to wait for more updates. Default: 1 s.
-
-
-	<tag><label id="rtable-max-settle-time">max settle time <m/time/</tag>
-	Specify a maximum value of the settle time. When a ROA table changes,
-	automatic <ref id="proto-rpki-reload" name="RPKI reload"> may be
-	triggered, after a short settle time. Maximum settle time is an upper
-	limit to the settle time from the initial ROA table change even if
-	there are consecutive updates gradually renewing the settle time.
-	Default: 20 s.
-
 	<tag><label id="rtable-gc-threshold">gc threshold <m/number/</tag>
 	Specify a minimum amount of removed networks that triggers a garbage
 	collection (GC) cycle. Default: 1000.
@@ -980,6 +965,15 @@ inherited from templates can be updated by new definitions.
 	<ref id="bgp-export-table" name="export table"> (for respective
 	direction). Default: on.
 
+	<tag><label id="rtable-min-settle-time">min settle time <m/time/</tag>
+	Minimum settle time is a delay from the last ROA table change to wait
+	for more updates before triggering automatic reload. Default: 1 s.
+
+	<tag><label id="rtable-min-settle-time">min settle time <m/time/</tag>
+        Maximum settle time is an upper limit to the settle time from the
+        initial ROA table change even if there are consecutive updates gradually
+        renewing the settle time.  Default: 20 s.
+
 	<tag><label id="proto-import-limit">import limit [<m/number/ | off ] [action warn | block | restart | disable]</tag>
 	Specify an import route limit (a maximum number of routes imported from
 	the protocol) and optionally the action to be taken when the limit is
diff --git a/nest/config.Y b/nest/config.Y
index 3ad6530d..ea7846e2 100644
--- a/nest/config.Y
+++ b/nest/config.Y
@@ -227,8 +227,6 @@ table_opt:
        cf_error("Trie option not supported for %s table", net_label[this_table->addr_type]);
      this_table->trie_used = $2;
    }
- | MIN SETTLE TIME expr_us { this_table->min_settle_time = $4; }
- | MAX SETTLE TIME expr_us { this_table->max_settle_time = $4; }
  | GC THRESHOLD expr { this_table->gc_threshold = $3; }
  | GC PERIOD expr_us { this_table->gc_period = (uint) $3; if ($3 > 3600 S_) cf_error("GC period must be at most 3600 s"); }
  | CORK THRESHOLD expr expr {
@@ -323,6 +321,8 @@ channel_item_:
  | RECEIVE LIMIT limit_spec { this_channel->rx_limit = $3; }
  | IMPORT LIMIT limit_spec { this_channel->in_limit = $3; }
  | EXPORT LIMIT limit_spec { this_channel->out_limit = $3; }
+ | MIN SETTLE TIME expr_us { this_channel->min_settle_time = $4; }
+ | MAX SETTLE TIME expr_us { this_channel->max_settle_time = $4; }
  | PREFERENCE expr { this_channel->preference = $2; check_u16($2); }
  | IMPORT KEEP FILTERED bool {
      if ($4)
diff --git a/nest/proto.c b/nest/proto.c
index d2374792..f4615c2f 100644
--- a/nest/proto.c
+++ b/nest/proto.c
@@ -55,6 +55,7 @@ static void channel_update_limit(struct channel *c, struct limit *l, int dir, st
 static void channel_reset_limit(struct channel *c, struct limit *l, int dir);
 static void channel_feed_end(struct channel *c);
 static void channel_export_stopped(struct rt_export_request *req);
+static void channel_check_stopped(struct channel *c);
 
 static inline int proto_is_done(struct proto *p)
 { return (p->proto_state == PS_DOWN) && proto_is_inactive(p); }
@@ -312,10 +313,19 @@ proto_remove_channels(struct proto *p)
     proto_remove_channel(p, c);
 }
 
+struct roa_subscription {
+  node roa_node;
+  timer t;
+  btime base_settle_time;		/* Start of settling interval */
+  struct channel *c;
+  struct rt_export_request req;
+};
+
 static void
-channel_roa_in_changed(void *_data)
+channel_roa_in_changed(struct timer *t)
 {
-  struct channel *c = _data;
+  struct roa_subscription *s = SKIP_BACK(struct roa_subscription, t, t);
+  struct channel *c = s->c;
   int active = !!c->reload_req.hook;
 
   CD(c, "Reload triggered by RPKI change%s", active ? " - already active" : "");
@@ -327,9 +337,11 @@ channel_roa_in_changed(void *_data)
 }
 
 static void
-channel_roa_out_changed(void *_data)
+channel_roa_out_changed(struct timer *t)
 {
-  struct channel *c = _data;
+  struct roa_subscription *s = SKIP_BACK(struct roa_subscription, t, t);
+  struct channel *c = s->c;
+
   CD(c, "Feeding triggered by RPKI change");
 
   c->refeed_pending = 1;
@@ -338,29 +350,56 @@ channel_roa_out_changed(void *_data)
     rt_stop_export(&c->out_req, channel_export_stopped);
 }
 
-/* Temporary code, subscriptions should be changed to resources */
-struct roa_subscription {
-  struct rt_subscription s;
-  node roa_node;
-};
+static void
+channel_export_one_roa(struct rt_export_request *req, const net_addr *net UNUSED, struct rt_pending_export *first)
+{
+  struct roa_subscription *s = SKIP_BACK(struct roa_subscription, req, req);
+
+  /* TODO: use the information about what roa has changed */
+
+  if (!tm_active(&s->t))
+  {
+    s->base_settle_time = current_time();
+    tm_start(&s->t, s->base_settle_time + s->c->min_settle_time);
+  }
+  else
+    tm_set(&s->t,
+	MIN(s->base_settle_time + s->c->max_settle_time,
+	    current_time() + s->c->min_settle_time));
+
+
+  rpe_mark_seen_all(req->hook, first, NULL);
+}
+
+static void
+channel_dump_roa_req(struct rt_export_request *req)
+{
+  struct roa_subscription *s = SKIP_BACK(struct roa_subscription, req, req);
+  struct channel *c = s->c;
+  rtable *tab = SKIP_BACK(rtable, exporter, req->hook->table);
+
+  debug("  Channel %s.%s ROA %s change notifier from table %s request %p\n",
+      c->proto->name, c->name,
+      (s->t.hook == channel_roa_in_changed) ? "import" : "export",
+      tab->name, req);
+}
 
 static int
 channel_roa_is_subscribed(struct channel *c, rtable *tab, int dir)
 {
-  void (*hook)(void *) =
+  void (*hook)(struct timer *) =
     dir ? channel_roa_in_changed : channel_roa_out_changed;
 
   struct roa_subscription *s;
   node *n;
 
   WALK_LIST2(s, n, c->roa_subscriptions, roa_node)
-    if ((s->s.tab == tab) && (s->s.event->hook == hook))
+    if ((s->req.hook->table == &tab->exporter) && (s->t.hook == hook))
       return 1;
 
   return 0;
 }
 
-
 static void
 channel_roa_subscribe(struct channel *c, rtable *tab, int dir)
 {
@@ -368,21 +407,40 @@ channel_roa_subscribe(struct channel *c, rtable *tab, int dir)
     return;
 
   struct roa_subscription *s = mb_allocz(c->proto->pool, sizeof(struct roa_subscription));
-  s->s.event = ev_new_init(c->proto->pool, dir ? channel_roa_in_changed : channel_roa_out_changed, c);
-  s->s.list = proto_work_list(c->proto);
 
-  rt_subscribe(tab, &s->s);
+  *s = (struct roa_subscription) {
+    .t = { .hook = dir ? channel_roa_in_changed : channel_roa_out_changed, },
+    .c = c,
+    .req = {
+      .name = mb_sprintf(c->proto->pool, "%s.%s.roa-%s.%s",
+	  c->proto->name, c->name, dir ? "in" : "out", tab->name),
+      .list = proto_work_list(c->proto),
+      .trace_routes = c->debug | c->proto->debug,
+      .dump_req = channel_dump_roa_req,
+      .export_one = channel_export_one_roa,
+    },
+  };
 
   add_tail(&c->roa_subscriptions, &s->roa_node);
+  rt_request_export(&tab->exporter, &s->req);
+}
+
+static void
+channel_roa_unsubscribed(struct rt_export_request *req)
+{
+  struct roa_subscription *s = SKIP_BACK(struct roa_subscription, req, req);
+  struct channel *c = s->c;
+
+  rem_node(&s->roa_node);
+  mb_free(s);
+  
+  channel_check_stopped(c);
 }
 
 static void
 channel_roa_unsubscribe(struct roa_subscription *s)
 {
-  rt_unsubscribe(&s->s);
-  rem_node(&s->roa_node);
-  rfree(s->s.event);
-  mb_free(s);
+  rt_stop_export(&s->req, channel_roa_unsubscribed);
 }
 
 static void
@@ -525,7 +583,7 @@ channel_check_stopped(struct channel *c)
   switch (c->channel_state)
   {
     case CS_STOP:
-      if (c->out_req.hook || c->in_req.hook)
+      if (!EMPTY_LIST(c->roa_subscriptions) || c->out_req.hook || c->in_req.hook)
 	return;
 
       channel_set_state(c, CS_DOWN);
@@ -533,7 +591,7 @@ channel_check_stopped(struct channel *c)
 
       break;
     case CS_PAUSE:
-      if (c->out_req.hook)
+      if (!EMPTY_LIST(c->roa_subscriptions) || c->out_req.hook)
 	return;
 
       channel_set_state(c, CS_START);
@@ -876,6 +934,9 @@ channel_config_new(const struct channel_class *cc, const char *name, uint net_ty
   cf->debug = new_config->channel_default_debug;
   cf->rpki_reload = 1;
 
+  cf->min_settle_time = 1 S;
+  cf->max_settle_time = 20 S;
+
   add_tail(&proto->channels, &cf->n);
 
   return cf;
@@ -956,6 +1017,22 @@ channel_reconfigure(struct channel *c, struct channel_config *cf)
   c->in_req.trace_routes = c->out_req.trace_routes = c->debug | c->proto->debug;
   c->rpki_reload = cf->rpki_reload;
 
+  if (	  (c->min_settle_time != cf->min_settle_time)
+       || (c->max_settle_time != cf->max_settle_time))
+  {
+    c->min_settle_time = cf->min_settle_time;
+    c->max_settle_time = cf->max_settle_time;
+
+    struct roa_subscription *s;
+    node *n;
+
+    WALK_LIST2(s, n, c->roa_subscriptions, roa_node)
+      if (tm_active(&s->t))
+	tm_set(&s->t,
+	    MIN(s->base_settle_time + c->max_settle_time,
+	      current_time() + c->min_settle_time));
+  }
+
   /* Execute channel-specific reconfigure hook */
   if (c->channel->reconfigure && !c->channel->reconfigure(c, cf, &import_changed, &export_changed))
     return 0;
diff --git a/nest/protocol.h b/nest/protocol.h
index 709d6415..a3eeb198 100644
--- a/nest/protocol.h
+++ b/nest/protocol.h
@@ -463,6 +463,9 @@ struct channel_config {
   struct channel_limit in_limit;	/* Limit for importing routes from protocol */
   struct channel_limit out_limit;	/* Limit for exporting routes to protocol */
 
+  btime min_settle_time;		/* Minimum settle time for ROA-induced reload */
+  btime max_settle_time;		/* Maximum settle time for ROA-induced reload */
+
   u8 net_type;				/* Routing table network type (NET_*), 0 for undefined */
   u8 ra_mode;				/* Mode of received route advertisements (RA_*) */
   u16 preference;			/* Default route preference */
@@ -490,6 +493,9 @@ struct channel {
   struct limit in_limit;		/* Input limit */
   struct limit out_limit;		/* Output limit */
 
+  btime min_settle_time;		/* Minimum settle time for ROA-induced reload */
+  btime max_settle_time;		/* Maximum settle time for ROA-induced reload */
+
   u8 limit_actions[PLD_MAX];		/* Limit actions enum */
   u8 limit_active;			/* Flags for active limits */
 
@@ -543,7 +549,7 @@ struct channel {
 
   struct rt_exporter *out_table;	/* Internal table for exported routes */
 
-  list roa_subscriptions;		/* List of active ROA table subscriptions based on filters roa_check() */
+  list roa_subscriptions;		/* List of active ROA table subscriptions based on filters' roa_check() calls */
 };
 
 #define RIK_REJECTED	1			/* Routes rejected in import filter are kept */
diff --git a/nest/rt-table.c b/nest/rt-table.c
index 72f85370..2fa736e2 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -82,9 +82,8 @@
  * will be re-validated later in this round anyway.
  *
  * The third mechanism is used for RPKI re-validation of IP routes and it is the
- * simplest. It is just a list of subscribers in src table, who are notified
- * when any change happened, but only after a settle time. Also, in RPKI case
- * the dst is not a table, but a channel, who refeeds routes through a filter.
+ * simplest. It is also an auxiliary export request belonging to the
+ * appropriate channel, triggering its reload/refeed timer after a settle time.
  */
 
 #undef LOCAL_DEBUG
@@ -134,7 +133,6 @@ static void rt_next_hop_update(rtable *tab);
 static inline void rt_next_hop_resolve_rte(rte *r);
 static inline void rt_flowspec_resolve_rte(rte *r, struct channel *c);
 static inline void rt_prune_table(rtable *tab);
-static inline void rt_schedule_notify(rtable *tab);
 static void rt_kick_prune_timer(rtable *tab);
 static void rt_feed_by_fib(void *);
 static void rt_feed_by_trie(void *);
@@ -1270,15 +1268,10 @@ rte_announce(rtable *tab, net *net, struct rte_storage *new, struct rte_storage
   if ((new == old) && (new_best == old_best))
     return;
 
-  if (new_best_valid || old_best_valid)
-  {
-    if (new_best_valid)
-      new_best->rte.sender->stats.pref++;
-    if (old_best_valid)
-      old_best->rte.sender->stats.pref--;
-  }
-
-  rt_schedule_notify(tab);
+  if (new_best_valid)
+    new_best->rte.sender->stats.pref++;
+  if (old_best_valid)
+    old_best->rte.sender->stats.pref--;
 
   if (EMPTY_LIST(tab->exporter.hooks) && EMPTY_LIST(tab->exporter.pending))
   {
@@ -2371,78 +2364,6 @@ rt_kick_prune_timer(rtable *tab)
 }
 
 
-static inline btime
-rt_settled_time(rtable *tab)
-{
-  ASSUME(tab->base_settle_time != 0);
-
-  return MIN(tab->last_rt_change + tab->config->min_settle_time,
-	     tab->base_settle_time + tab->config->max_settle_time);
-}
-
-static void
-rt_settle_timer(timer *t)
-{
-  rtable *tab = t->data;
-
-  if (!tab->base_settle_time)
-    return;
-
-  btime settled_time = rt_settled_time(tab);
-  if (current_time() < settled_time)
-  {
-    tm_set(tab->settle_timer, settled_time);
-    return;
-  }
-
-  /* Settled */
-  tab->base_settle_time = 0;
-
-  struct rt_subscription *s;
-  WALK_LIST(s, tab->subscribers)
-    ev_send(s->list, s->event);
-}
-
-static void
-rt_kick_settle_timer(rtable *tab)
-{
-  tab->base_settle_time = current_time();
-
-  if (!tab->settle_timer)
-    tab->settle_timer = tm_new_init(tab->rp, rt_settle_timer, tab, 0, 0);
-
-  if (!tm_active(tab->settle_timer))
-    tm_set(tab->settle_timer, rt_settled_time(tab));
-}
-
-static inline void
-rt_schedule_notify(rtable *tab)
-{
-  if (EMPTY_LIST(tab->subscribers))
-    return;
-
-  if (tab->base_settle_time)
-    return;
-
-  rt_kick_settle_timer(tab);
-}
-
-void
-rt_subscribe(rtable *tab, struct rt_subscription *s)
-{
-  s->tab = tab;
-  rt_lock_table(tab);
-  DBG("rt_subscribe(%s)\n", tab->name);
-  add_tail(&tab->subscribers, &s->n);
-}
-
-void
-rt_unsubscribe(struct rt_subscription *s)
-{
-  rem_node(&s->n);
-  rt_unlock_table(s->tab);
-}
-
 static void
 rt_flowspec_export_one(struct rt_export_request *req, const net_addr *net, struct rt_pending_export *first)
 {
@@ -2580,7 +2501,6 @@ rt_free(resource *_r)
   fib_free(&r->fib);
   hmap_free(&r->id_map);
   rfree(r->rt_event);
-  rfree(r->settle_timer);
   mb_free(r);
   */
 }
@@ -2642,8 +2562,6 @@ rt_setup(pool *pp, struct rtable_config *cf)
   hmap_init(&t->id_map, p, 1024);
   hmap_set(&t->id_map, 0);
 
-  init_list(&t->subscribers);
-
   t->rt_event = ev_new_init(p, rt_event, t);
   t->uncork_event = ev_new_init(p, rt_uncork_event, t);
   t->prune_timer = tm_new_init(p, rt_prune_timer, t, 0, 0);
@@ -3629,8 +3547,6 @@ rt_new_table(struct symbol *s, uint addr_type)
   c->addr_type = addr_type;
   c->gc_threshold = 1000;
   c->gc_period = (uint) -1;	/* set in rt_postconfig() */
-  c->min_settle_time = 1 S;
-  c->max_settle_time = 20 S;
   c->cork_threshold.low = 128;
   c->cork_threshold.high = 512;
   c->debug = new_config->table_debug;
diff --git a/nest/rt.h b/nest/rt.h
index 01e444f6..c5948512 100644
--- a/nest/rt.h
+++ b/nest/rt.h
@@ -61,8 +61,6 @@ struct rtable_config {
   byte sorted;				/* Routes of network are sorted according to rte_better() */
   byte trie_used;			/* Rtable has attached trie */
   byte debug;				/* Whether to log */
-  btime min_settle_time;		/* Minimum settle time for notifications */
-  btime max_settle_time;		/* Maximum settle time for notifications */
   btime export_settle_time;		/* Delay before exports are announced */
   struct rt_cork_threshold cork_threshold;	/* Cork threshold values */
 };
@@ -112,7 +110,6 @@ typedef struct rtable {
   struct event *uncork_event;		/* Called when uncork happens */
   struct timer *prune_timer;		/* Timer for periodic pruning / GC */
   btime last_rt_change;			/* Last time when route changed */
-  btime base_settle_time;		/* Start time of rtable settling interval */
   btime gc_time;			/* Time of last GC */
   uint gc_counter;			/* Number of operations since last GC */
   byte prune_state;			/* Table prune state, 1 -> scheduled, 2-> running */
@@ -130,18 +127,9 @@ typedef struct rtable {
   u32 trie_old_lock_count;		/* Old prefix trie locked by walks */
   struct tbf rl_pipe;			/* Rate limiting token buffer for pipe collisions */
 
-  list subscribers;			/* Subscribers for notifications */
-  struct timer *settle_timer;		/* Settle time for notifications */
   struct f_trie *flowspec_trie;		/* Trie for evaluation of flowspec notifications */
 } rtable;
 
-struct rt_subscription {
-  node n;
-  rtable *tab;
-  event *event;
-  event_list *list;
-};
-
 extern struct rt_cork {
   _Atomic uint active;
   event_list queue;
@@ -461,8 +449,6 @@ void rt_lock_table(rtable *);
 void rt_unlock_table(rtable *);
 struct f_trie * rt_lock_trie(rtable *tab);
 void rt_unlock_trie(rtable *tab, struct f_trie *trie);
-void rt_subscribe(rtable *tab, struct rt_subscription *s);
-void rt_unsubscribe(struct rt_subscription *s);
 void rt_flowspec_link(rtable *src, rtable *dst);
 void rt_flowspec_unlink(rtable *src, rtable *dst);
 rtable *rt_setup(pool *, struct rtable_config *);