I like this idea, in theory. But it feels it aligns with the trend to apply overkill to lots of real-world security scenarios nowadays.

I'm sure that in some situations this kind of approach would be up there on the drawing board. Lots of money pitted against unknown trust, or military projects (the super-super expensive kind) immediately come to mind.

But even then, in the theoretical "high risk / untrustworthiness" scenario, surely one payout event wouldn't provide enough confidence to predict similar behavior in the future, and causing multiple events is suspicious enough you wouldn't be able to rule out whether the adversary was paying out each time because they'd caught on.