A retry routine is essentially an extension of the mainline routine. When an error occurs, the system passes control to your recovery routine, which can then request the system to pass control back to the mainline routine to resume processing. That portion of the mainline that gets control back is referred to as the retry routine. When the retry routine gets control, it is as if the mainline routine branched there after encountering the error; to the mainline routine, it appears as though the error never occurred.
The retry routine does whatever processing your mainline routine would continue doing at that point.
Once the retry routine is running, if another error occurs, the system again passes control to your recovery routine, just as it did when the mainline routine encountered an error.