What's the downside of very small checkpoint retention time?

I've been reading the 11.2g Streams documentation and while I see what checkpoint_retention_time is (how long checkpoints remain and thus how close to required_checkpoint_scn the system will auto-advance first_scn and start_scn to), I remain unclear on what the effects on the system are and what the downside is of a very short retention time.

So what exactly is the downside of keeping checkpoint_retention_time at a day or less? Will it cause problems if streams replication is down for longer than the retention time? (Say I have retention set to 1 day and one of the other database instances is down for longer than a day -- will LCRs be "lost" due to the checkpoint being purged? Or will checkpoints not be purged if required_checkpoint_scn is not advancing? (As I understand things, it won't advance if replication is down to another instance)) Are there other problems?