Addition to the 'Matrix of possible variable combinations' of ID handling during import

The matrix lists 6 cases of possible variable combinations, to explain how the module handles IDs during imports.
One of the assumptions that this matrix is built upon is:

When key provided is false, then provided key already exists can only be false

This makes sense, only that this assumption doesn't take into account the alternative keys (take a look at "Node title should not be an alt key of nid").

I have an keyless import, where I happen to have two documents with the same title. Importing the data creates confusions and inconsistencies in the database, leading to zombie nodes - they appear in the node table, but not in the revision table, or vice-versa. This makes it impossible to view them, or even delete them, not to mention taxonomy or files attribution, without direct manipulation of the database.

What should happen then, in the 7th (undocumented) case - When key provided is false and alt key already exists?