As the article mentions, the margins on the restaurant business are slim enough that the wage pool is effectively fixed - it's not a hypothetical scenario, it's a practical constraint.
If you need to pay somebody $2 more an hour the money can come from more places. Increase a drink price, decrease tomato quality, steam clean every 3 weeks instead of 2, wash the bathroom yourself instead of hiring, call the newspaper and ask for a reduced rate, etc etc etc.
Of course, you are also assuming that decisions like these (aside from the newspaper one) don't affect the revenue of the restaurant in the long run.
As he states, "Theoretically, it seems like there would be ways to free up more money for labor, and there are some methods that help. In the end, though, restaurant-going is so prevelant that the expectations of the market — for price, quality, ambience, and so forth — are pretty well set in every dimension. And the cost of meeting those expectations is pretty well set, too."
Which means his entire experiment is how to increase revenue that can be directly allocated to wages without decreasing the quality of his tomatoes or the cleanliness of his carpets.
Juggling and heroics make the variables of a business much more dependent on one another, not less.