Have you considered allowing user(s) to specify the segmentation used in Live Aggregate Projections?
Question from The shortest path to Vertica: Best practices for data warehouse migration & ETL@marcothesane @Maurizio
I suppose someone like @ChuckB can answer this to more detail than me - but:
If you remember what an optimised flow of data looks like in a GROUP BY operation: the rows to be grouped are all on the same node to avoid global resegmentation - which will define how to segment the new projection, and they are sorted by the grouping columns, to allow for a pipelined group-by, and this will define how the new projection will be sorted. Any attempt to change that will invariably cost you a higher price than any benefits you might get out of it ...
In addition, we want a lot of operations on LAPs to be node-local, including how aggregation gets completed.