I would be surprised if this even works against the average 911-era terrorist. Stochastic enforcement with 10% success is good enough to stop well to-do people from shoplifting (small benefit, large expected cost) and to catch repeat offenders, but the success criterion for a suicide bomber is to explode and die! Why wouldn’t they just YOLO it?
I suppose it might be different if, say, you have a dozen checkpoints and only one dog, and the bombers could probably avoid a single checkpoint without getting caught sneaking over a fence or something, but if the bombers jump a dozen fences they’re likely to be caught, and if they go through all the checkpoints the one with the real dog will get them.
The actual reason is probably the boring fact that political functions (laws) almost never take macroeconomic metrics as input, except population. Some take individual measures like “the sale price of a good” or “a company’s income”. It would be more idiomatic to apportion voting power to a house of district governors based on taxes contributed to the federal budget.
As to why that’s never happened. I dunno. I think basically any representation apportionment method except population or arbitrary clustering (or like, tradition, I guess? but there’s no tradition for monetary contribution) is pretty much taboo, and something something Chesterton’s Fence (though population-based voting has started to run into the Old People Voting Themselves Infinite Money exploit in the west).