One calculation, two questions
“How is this person related to me?” and “How are these two people related?” are the same computation with a different starting point. The archive runs one engine for both, so the two answers can never disagree.
One calculation answers both 'how is this person related to me?' and 'how are these two people related?'