Also, Elasticsearch indexes string dates, not numeric dates. So users would not be able to take advantage of both ES and N1QL because the two engines would have conflicting formats for dates stored in documents.
An array based date index would offer flexibility for N1QL to further develop query abilities when it came to dates. You could do range scans at different levels (year, month, date, etc.) and potentially get aggregate data utilizing CB's grouped reduce queries.