Hey thanks for putting some thoughts into my problem. I tried both boundary distance and volume based methods already. The former turned out to give a slightly better overset interface. What I have also tried:

Different boundary types for the faces at the interface/overlap between the two component bodies
Different grid priorities, manual cut-controls and specific definition of overlapping boundaries
Modified donor search and modified hole cutting parameters

Especially specifying overlapping and non-overlapping boundaries improved the overset interface quite a lot, but didn't remove all the orphan cells. As you can see in the figures below, I'm still left with orphan cells all around the overlapping region (marked in red). Like this, the setup is pretty unstable and diverges after a few iterations.