This theory has to explain why the experts, the people paid by campaigns, and the campaigns themseleves, spend so much $ on polls maybe the experts are just beltway bandits ?

The column isn’t saying polls are meaningless! But it’s saying that the perfectly normal error in polls—the amount of error we see every single election—is enough to result in “Harris wins every battleground” *or* “Trump wins every battleground”. (And that’s information campaigns need to know!)