Amount of non-purgeable archive logs for Streams

I have bi-directional table-level Streams replication between a Solaris 10.2.0.3 database and a Linux 10.2.0.4 database.  I have this configuration implemented in 3 dev./test database pairs.  On one of the Solaris DBs, I see this:

SQL> SELECT MIN(FIRST_TIME), MAX(FIRST_TIME)
  2    FROM DBA_REGISTERED_ARCHIVED_LOG r, DBA_CAPTURE c
  3    WHERE r.CONSUMER_NAME =  c.CAPTURE_NAME AND
  4        r.NEXT_SCN      >= c.REQUIRED_CHECKPOINT_SCN AND
  5        r.PURGEABLE='NO';

MIN(FIRST_TIME)      MAX(FIRST_TIME)
-------------------- --------------------
08-MAR-2010 22:01:46 25-MAR-2010 19:36:39

I have checkpoint_retention_time set to 7 and _checkpoint_frequency set to 1000. 

Why do I need 17 days' worth of archive logs?  When I add a new table to an existing Streams configuration, I see in the alert log that it’s mining all of those past archive logs before it is done.  This has been taking around 30 minutes in my development environment.  I can't imagine how long it will take when I implement this in production.  How can I reduce this?