Far transfer, meaning borrowing across distant fields, is both powerful and risky. The most productive scientific analogies come from comparing a problem to something in a remote field, and breadth of experience predicts this kind of innovation better than depth in one specialty. But the same mechanism fails hard when the apparent structural match doesn't actually hold.
A sharp diagnostic: does the source domain involve intelligent agents where the target doesn't, or vice versa? The disease analogy for cybersecurity works for layered defense but fails for modeling attackers, because pathogens mutate blindly while hackers reason strategically about your defenses. Agency changes system behavior, and analogies that ignore that asymmetry mislead. Practice stripping a problem to its relational skeleton first. See Types of Reasoning.