Our 19 hour job was reduced to about 30 minutes using this technique.
We had to import 110 million rows into a SQL Server table containing 5 ints during 4 import sessions. There was a clustered index over 4 of those ints.
We used the System.Data.SqlClient.SqlBulkCopy class to do the import (class mirroring the sql server bcp utility). The initial import took ~19 hours. On a new table we removed the clustered index and inserted all the rows again. This took about 12 minutes. Applying the clustered index took about 20 minutes.