Unfortunately I don't think we currently have a workflow that quite accommodates this. If you are able to use a CSV import with the match-and-update option, then your digital object should not be affected if you leave the DO columns blank in your update CSV import. However, with the delete-and-replace method, we don't currently have a way of holding the digital object and re-attaching it to a replaced description.
We did attempt to do something like this with the match and update option for CSV imports - see this ticket:
However, this was to solve a particular use case during a 2.3 migration - i.e. prior to the current update/replace options. As you can see from the discussion on the ticket linked above, it's not very useful with the more recent changes in 2.4, because you can achieve the same thing by simply not including new digital object information in your CSV - AtoM will interpret blank rows not as a command to overwrite existing data with NULL values, but as a cue to skip the column, leave the current data untouched, and continue with the import.
I think we would have to refactor the job significantly to support this workflow, so it could track the unattached digitial object after deletion, and re-attach it correctly. However, there would also be a lot of edge cases. As a simple example: If you had originally attached it to a series, and your replacement file includes no series at all, what is AtoM to do with the object?
In most cases for now, it may be better to re-import or upload the digital objects post replacement.
Regards,