In case anyone cares:

Found that excessive amount of logs were being used during refresh of a
4.5 million row summary table with 3 indexes, which seemed suspicious.
Research showed that a previously refreshed summary table logs the
deletes in addition to the inserts. All of this work is done in one
unit of work, thus secondary log use becomes excessive. If there are
indexes on the summary table, they cause even more logs to be created.
Here are the results of log space used on a refresh with a 4.5 million
row summary table with 3 indexes:

In Bytes:

Refresh with 3 indexes = 8494062221
Drop indexes, refresh, create indexes = 4200950270
Drop/Recreate summary table, refresh, create indexes = 1474991094

To work around the amount of logging taking place, we decided to drop
and re-create the summary table before running the refresh and create
indexes after the refresh.

The order of events are:

1. drop summary table
2. create summary table
3. refresh summary table
4. create indexes
5. runstats