Processed changesets are being reindexed when subsequent change sets are processed. This has the effect of slowing down the indexing process considerably.

This only applies to the Lucene (ie. text searching) indexing - the actual change set information itself is not reprocessed.