I think that’s because that only would be truly beneficial once you have huge tables, and you really wouldn’t want your database to suddenly decide “let me store this table order by its first column, as that seems to be faster for the queries I run now” if that choice leads to degraded performance for hours while the database reorganises tables, creates indexes, etc.

And it can get even worse; that decision might make the queries it runs _now_ go from “more than fast enough” to “way more than fast enough”, but also slow down the queries you run infrequently from “will do” to “glacial”.

I think we would be better of with a database that can easily generate a report showing queries run in the last months, with suggested schema changes and their estimated impact on performance. SQL Server has the query analyzer, but I think that could be taken much farther and be made easier to use.

Edit: I see that, while I wrote the above, jws wrote that Digital's RDB had something like that.