Snapshots are a simple way of extracting data from a system, since no archivability checks are performed. So why not use this function to extract the entire data set of a system, particularly if the system will be shut down anyway?
The answer has to do with the nature of snapshots. In contrast to standard archiving objects designed for archiving business-complete data, snapshots are intended for archiving data from business processes that are still open for any reason. However, due to the special character of snapshots (remember: snapshots archive business-incomplete data that does not have a final status, such as “complete”) it is not possible to calculate the expiration date as for business-complete data.
Therefore, in the ILM store snapshots are stored with an “unknown” expiration date set. If you need to destroy a snapshot, for example, because a newer snapshot exists, you can do so by setting the expiration date manually to a specific date in the future. However, this will always be a manual process reserved for the exceptional case of snapshots or other data without time relevance. Therefore, archiving business-complete data as snapshots does not make sense, although it is technically possible.