Reindexing is typically a lot more resource intensive as all documents need to be reprocessed. The shrink index API combines the existing segments without reprocessing, so is a lot faster and results in less CPU and disk I/O usage. The bigger the index the bigger the difference.