* No crash/external event. Coke goes up. Pepsi goes up more due to health products. They're out money on Coke, but make enough on Pepsi to cover it and then some.
* Crash/external event. Both drop. Pepsi drops less because of the health products. They lose some on Pepsi, but because Coke dropped more, they make more off the Coke short than was lost on Pepsi. They still come out ahead.
* If the health products have no effect, then the gain or drop will be similar, giving roughly net-zero cost/gain.
The real risk, as I see it, is that the correlation breaks the other way, and Pepsi drops while Coke goes up. That's the price you pay trying to make a bet on something - the fund is safe from external events, but you can't hedge against being dead wrong and still make a profit.
In the canonical example, if they think that Pepsi will perform well because of some new health products, then they will go long Pepsi and short Coke. The idea is that all the other events they haven't looked at: a crash, currency shifts, people decide sugar is bad for you, NYC soda ban etc will hit both companies equally hard. The only thing they want to bet on is their single hypothesis.
They can't always do this cleanly, but to the extent they can, they diminish some of the risk around the correlation of asset classes.