Fact %2 sampling Insert with BIs -
			16427326 rows created.
            Elapsed: 00:22:15.34

without BIs - 
            16427326 rows created.
            Elapsed: 00:00:57.99

Fact %10 sampling Insert with BIs -
            83784620 rows created.
            Elapsed: 00:54:08.11

without BIs - 
            83784620 rows created.
            Elapsed: 00:04:48.92

So if BIs were chosen to be set unusable before ETL steps and rebuilded serially afterwards;

BI rebuilding on another fact table's reference column with higher cardinality -
            Elapsed: 01:02:53.26

BI rebuilding on one dimension table's reference column with low cardinality -
            Elapsed: 00:10:25.68

BI rebuilding on one dimension table's reference column with low cardinality -
            Elapsed: 00:09:48.31

BI rebuilding on one dimension table's reference column with low cardinality -			
            Elapsed: 00:09:56.63

BI rebuilding on one dimension table's reference column with low cardinality -
            Elapsed: 00:10:37.18

BI rebuilding on one dimension table's reference column with low cardinality -
            Elapsed: 00:10:40.60

BI rebuilding on one dimension table's reference column with low cardinality -
            Elapsed: 00:10:16.28

The test schema had fact tables this sized;
	FACT TABLE 1 SUBPARTITION 	97,0 GB
	FACT TABLE 2 SUBPARTITION 	14,3 GB
And each low cardinality bitmap indexes were sized around 13,0 GB.