I think I understand-ish why this is happening, and have a fix that makes it work, but I'm not sure if there are any side effects.
First the fix:

--- plugins/pager/pager.c	2015-02-26 20:32:17.000000000 +0000
+++ fixed/plugins/pager/pager.c	2015-05-31 16:23:35.682554450 +0100
@@ -320,8 +320,8 @@
   if (G_UNLIKELY (plugin->pager != NULL))
     {
       gtk_widget_destroy (GTK_WIDGET (plugin->pager));
-      wnck_screen_force_update (plugin->wnck_screen);
     }
+  wnck_screen_force_update (plugin->wnck_screen);
 
   mode = xfce_panel_plugin_get_mode (XFCE_PANEL_PLUGIN (plugin));
   orientation =

---------------------

Now some explanation; I added a lot of debug in libwnck

wnck_pager_set_layout_hint: Entry for 0x26bb1e0
_wnck_pager_set_screen: entry for 0x26bb1e0
_wnck_pager_set_screen: Exit for !gtk_widget_has_screen
wnck_pager_set_layout_hint: pager->priv->screen==NULL
libpager-Message: Setting the pager rows returned false. Maybe the setting is not applied.
_wnck_pager_set_screen: entry for 0x26bb1e0
wnck_pager_set_layout_hint: Entry for 0x26bb1e0
wnck_screen_try_set_workspace_layout entry: screen=0x26938b0 current_token=0 rows=3 columns=0
wnck_pager_set_layout_hint: bottom 1
_wnck_pager_set_screen: Exit bottom
wnck_pager_set_layout_hint: Entry for 0x26bb620
_wnck_pager_set_screen: entry for 0x26bb620
_wnck_pager_set_screen: Exit for !gtk_widget_has_screen
wnck_pager_set_layout_hint: pager->priv->screen==NULL
libpager-Message: Setting the pager rows returned false. Maybe the setting is not applied.
_wnck_pager_set_screen: entry for 0x26bb620
wnck_pager_set_layout_hint: Entry for 0x26bb620
wnck_screen_try_set_workspace_layout entry: screen=0x26938b0 current_token=0 rows=3 columns=0
wnck_pager_set_layout_hint: bottom 0
_wnck_pager_set_screen: in !set_layout_hint branch, before get_workspace_layout n_rows=3
!!! _wnck_pager_set_screen: in !set_layout_hint branch, after get_workspace_layout n_rows=1
_wnck_pager_set_screen: Exit bottom
update_workspace_layout on 0x26938b0
!!! update_workspace_layout on 0x26938b0: num_workspaces=9 rows=3 cols=0
update_workspace_layout on 0x26938b0

Note the two lines with !!!
What I think happens is that the 1st pager instance works OK, but for the 2nd pager we end up in _wnck_pager_set_screen   and it has:
  if (!wnck_pager_set_layout_hint (pager))
    {
      _WnckLayoutOrientation orientation;

      /* we couldn't set the layout on the screen. This means someone else owns
       * it. Let's at least show the correct layout. */
      _wnck_screen_get_workspace_layout (pager->priv->screen,
                                         &orientation,
                                         &pager->priv->n_rows,
                                         NULL, NULL);


that should ensure that the 2nd pager has the same layout, but it doesn't work because as from that debug we see:

_wnck_pager_set_screen: in !set_layout_hint branch, after get_workspace_layout n_rows=1

so _wnck_screen_get_workspace_layout is getting the wrong answer - why?
Well I think it's set by update_workspace_layout, and we only see that called
later in the debug; it's normally called on idle - see update_idle.
But I guess there's no idle between the two pagers, so we get to the second pager and haven't setup it's screen's row data yet.
I'm not sure if wnck3 is any better at this; but is there any downside to just doing that force_update ?

Dave