Performance of has_child query for document with large text

I have this situation, documents have some fields with lots of text that does not change often, and a lot of other fields (used for filtering, faceting, etc that changes often).

Now whenever I need to change one of these properties, I need to reindex the whole document in ES, thus asking ES to parse again the same text data.

One possible solution is storing the fields with lots of text in child object, and issue query with has_child (ES 1.7). With such scenario I'll reindex the text only when the text changes.

I'm curious if anyone has tested loss of performance when it comes the time to query adopting this solution.

Gian Maria.