Big fan of the transparent data-driven de-duping ("Is this a duplicate? The ID of the person to merge into" field on each record).
Since the merge process starts by relating record IDs, it seems possible to leverage that into batch deduping by loading a two column file containing "current record ID of a duplicate" and "the ID of the person to merge into".
I can see where you might prefer in the long-run to develop internal tools for matching and batch de-duping. But the capability to export, then use external matching tools, and then load in the results would be great to have in the meantime.