Some thoughts;
Some time ago I complained about an import that had brought in many duplicate profiles which resulted in many tree matches which had to be dealt with. I expressed the opinion that Geni should allow it. One of the curators, I don't remember who, pointed out that the import had brought in several10's of thousands of profiles, and that for the WFT to grow we should welcome imports, even if it means some work reconciling the duplicates, in order for the WFT to become more complete. There is some truth to that, and it obviously depends on the quality of the data and the ratio of new profiles to duplicate profiles to be able to evaluate its usefulness.
I don't think Job was suggesting not to merge the resultant duplicates. I think he was suggesting not to rush into doing merges while the import was still running, as it might affect the algorithms and allow the import of even more duplicates. But after the import has finished I think it is critical the the duplicate profiles be merged.
The suggestion that merges should routinely be rejected is a mistake, because leaving duplicates in the WTF without attempting to fix them, degrades our common end product. Ideally the user who ran the import would fix all the dups, but unfortunately that can't be relied on. I for one hate having tree merges associated with any profiles I manager or follow and attempt to reconcile the issues If I am capable.
Bernard