It is considered a best practice to regularly commit batches during non-peak hours. Uncommitted batches are stored only in the first mass storage copy of the Disk Group; if this disk was to fail, these batches would be lost.
When batches are committed, documents in the batches are copied to the secondary and tertiary copies of the Disk Group. If one of these Disk Groups was to fail, the data could be recovered from another copy of the Disk Group.