Yes I believe it is supposed to be like that. Everything, versions of pages, document etc that you put in index is a document in the lucene index.

Nested is document, but it is special one. It can't be retrivied from index without "parent" doc (http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/mapping-nested-type.html)

So probably there is no a lot of sense to take them into account to limit index capacity.

As for me, more reasonable is to limit index physical size instead of documents count.

Yes I agree, they have problem with ecommerce sites were document count is a big issue. Hopefully they will come up with a new plan since it is a bit tricky to keep the count down on larger websites.

It also appears to count page types which have been excluded (via Conventions.ForInstancesOf). This leave me with a lot less that 10,000 documents in the the index but I still reach the 10,000 limit when using the schedlued job to re-index. Could anyone from EPiServer confirm this?