> Hmm, you realise Alvaro is working on MinMax indexes in this release?
> They are very efficient with regard to index inserts and specially
> designed for use on large tables.
>
> Prior work by Heikki on Grouped Item Tuples was a way of reducing the
> size of indexes, yet still allowing uniqueness checks. That is
> implemented in SQLServer already and is very useful.
Reading the implementation of those features, I don't think they can help in the cases handled by the index types I mentioned (insertions of random values in big tables).