Different subpartitioning schemes for mviews and PCT-refresh

Question

It appears that I can have a table RANGE-partitioned by some field (DATE) and LIST-subpartitioned by another field (let's call it subfield2). Then I create materialized view (mview) based on that table, RANGE-partitioned by the same DATE field, but subpartitioned by another field (subfield1, logically parent of subfield2). Then I create another mview, again RANGE-partitioned by the same DATE field, but not subpartitioned at all.
PCT-refresh (dbms_mview.refresh('mview_name', 'P')) works fine, producing correct results. But I wonder if subpartitioning is taken into account when refreshing mviews. My suspicion is that only primary partitioning is taken into account, and primary partitions are refreshed as a whole, even if only one or few subpartitions have been modified in a table. Are my assumptions correct?
Another question - can I preserve partitions in mview if dropping them in underlying table? Table stores raw data, and legacy partitions may be dropped to free up some space. But mviews which store data aggregated on some levels may need to preserve data either longer than in raw table, or forever. I could not find a way of achieving this - when I drop a partition from a table on which mviews are built, and refresh mviews - data from dropped partitions of a table is gone from mviews, even though partitions of mviews are not deleted.

Answer 1

It appears that I can have a table RANGE-partitioned by some field (DATE) and LIST-subpartitioned by another field (let's call it subfield2). Then I create materialized view (mview) based on that table, RANGE-partitioned by the same DATE field, but subpartitioned by another field (subfield1, logically parent of subfield2). Then I create another mview, again RANGE-partitioned by the same DATE field, but not subpartitioned at all.
PCT-refresh (dbms_mview.refresh('mview_name', 'P')) works fine, producing correct results. But I wonder if subpartitioning is taken into account when refreshing mviews. My suspicion is that only primary partitioning is taken into account, and primary partitions are refreshed as a whole, even if only one or few subpartitions have been modified in a table. Are my assumptions correct?
Another question - can I preserve partitions in mview if dropping them in underlying table? Table stores raw data, and legacy partitions may be dropped to free up some space. But mviews which store data aggregated on some levels may need to preserve data either longer than in raw table, or forever. I could not find a way of achieving this - when I drop a partition from a table on which mviews are built, and refresh mviews - data from dropped partitions of a table is gone from mviews, even though partitions of mviews are not deleted.