I thought about that crosstalk or more like: "how make noise between "to small" transistors useful?". My gut says we can.
We might have to deal with incorrect results that require repeating the operation 3+ times. The usefulness (speed) would have to make up for that.
Are fluid or fluid-like substances as an interconnect viable? Idea being you just chuck your discrete components into some interconnect goo and magic does the rest.