duplicate nodes for XML Auto

It's no need to use join when you using in-column sub-queries. 

The first query could like this:

select cust.cust_Id,    id,    timestamp,    uid,    [user],    visible,    [version],    changeset,    [action] ,
  (SELECT distinct ref  FROM [nd] as nd where nd.cust_Id = cust.cust_Id and tag.cust_Id  = ND.cust_Id FOR XML raw('nd'), type ),
 k , v  
from cust as cust
--join [dbo].[nd] as nd
--on ND.cust_Id = cust.cust_Id
join cust_tags as tag
on tag.cust_Id = cust.cust_Id
where cust.cust_Id=23165
for xml   auto

and the second one would be this

select cust.cust_Id,    id,    timestamp,    uid,    [user],    visible,    [version],    changeset,    [action], 
(SELECT ref  FROM [nd] as nd where nd.cust_Id = cust.cust_Id FOR XML raw('nd'), type
 )    ,
 (SELECT k , v  FROM cust_tags as tag where tag.cust_Id = cust.cust_Id FOR XML raw('tag'), TYPE
 )
from cust as cust
--join cust_tags as tag
--on tag.cust_Id = cust.cust_Id
--join [GIS].[dbo].[nd] as nd
--on ND.cust_Id = cust.cust_Id
for xml auto