We have two very different kind of hybrid queries:

• projection only hybrid queries (for instance metadata hybrid queries - or to fetch non projectable index data)
• hybrid queries that need to sort the result set in memory (not on the index side!)

The former perform almost as well as indexed queries, since they do just an extra effort in loading some data from the cache.
The latter perform very badly and they need to have the max result unbounded.

With this issue we want to differentiate the two cases, somehow (subclasses, flag, ...).
And apply the max result to the Lucene queries in the first case.

This should improve the projection only hybrid queries performance.