The very simplest way would be to just change pageForIAItem() to always return an empty string.

The next simplest would be to add a checkbox ("Ignore duplicates") to the UI that pageForIAItem() checks and if true return the empty string.

With the latter approach the solution could eventually be expanded so that the checkbox is hidden by default, and when pageForIAItem() detects a duplicate it sends the user to a form where that checkbox is made visible (and possibly checked by default). That would give us some protection against unintended dupes, but may not be worth the effort in practice.

The two first variants should be doable within roughly "straightforward bugfix" amounts of time expended. The latter probably isn't prohibitively hard to do, but that would depend on what facilities are already available and would in any case take non-trivially more effort than the two dumbest approaches. But maybe it could piggyback on the functionality that returns information about the duplicate file to the user?

In any case, the dupe check has as a fundamental assumption that the existence of duplicates will be an exception state. After the bulk-upload of significant portions of IAs works that assumption no longer holds true: not having a dupe is now the exception, and having a dupe is the expected state. In other words, so long as this check is in place as-is the tool is functionally broken from an end-user perspective.