Union/Range Partitioning and Aggregation Symmetry

It often happens that facts are not a single table, but a union of tables. Why such practice is likely to remain even after range partitioning and roll-in/roll-out functionality becomes available in DB2? How should we approach the MQT definitions over union-partitioned fact tables? How can we ensure that the optimizer will rewrite the queries to use the MQTs? How should we deal with the opposite problem, where a yearlong base table is covered by 12 period MQTs?