This fetch command is useful for SELECT queries on large data sets. First, it runs a query to determine the number of rows being returned, then preallocates the space necessary for the input query's result, then runs the main query, fetching the results in batches.
This allows pulling much more information from a database in a single query than the built-in methods (by using batches to avoid using too much of the Java heap) and does this much faster (by preallocating space for the output).

Note for r2013a: This functionality is now built into the 'fetch' command in the Database Toolbox. See the documentation for that command.