Maybe there is such a thing as a free appetizer

This week, Freakonomics Radio broadcast their take on so-called "free appetizers" - bread, chips, etc. that many restaurants supposedly give out for "free" with their meals.
Other supposed freebies include free soft drink refills, free unlimited salad bar with your meal, and even free french fries (and other sides!) at some casual restaurants. A listener contacted Dubner and asked, "Why do restaurants do this?" The program puts forth 6 hypotheses of varying plausibility:

1. Increase turnover - people who fill up on bread eat their meal and leave, allowing the owner to clear the table and turn it over. People who stay for dessert linger too long. The experts in the podcast don't buy this one, and neither do I. Desserts may be low in menu price, but they typically have higher profit margins than entrees. People who stay longer are often also getting other high-margin items like coffee or alcohol. 
2. Reduce complaints - Hungry people waiting for their order get cranky and may complain or leave. The broadcast also suggests that, for behavioral reasons, "free" is much more effective at this than just "cheap." This probably doesn't explain a lot of it, and I don't really buy the "behavioral" angle on this explanation, but it's a plausible contributor.
3. Exploit complentarities with higher-margin items - A basket of salty chips or bread makes people buy other things that "go with" that item. The usual suspect here is drinks, which fits with the "loss leader" strategy. I would also point out that at many Mexican restaurants, about half of the available appetizers are direct complements to chips: cheese or bean dip, fundito, and guacamole. 
4. Expectations and history - At some point a firm saw it in its strategic advantage to give stuff out for free, or to charge a fixed-price. Other firms followed suit, consumers developed expectations around receiving free stuff to snack on, and the rest is history. 

Finally, they propose that the "free" apps are not free at all, but rather that their cost is included in the price. This leads them to the question, "why not let consumers choose (or not choose) the appetizer they want, and price them separately?" This is really the key question, and the one that institutional analysis has a good answer for. Bottom line: Transactions costs matter.
To answer the first part of the question, then from the firm's perspective there is a value in knowing that all consumers that place an order will be ordering the same thing. If economies of scale matter, then making the same thing 400 times at predictable intervals each night will be cheaper than making 8 different things an average of 50 times each at unpredictable intervals. Consumers also then have to make their choice, which itself takes time, energy, and as Yoram Bauman comically points out, "choices are bad - really bad." In some contexts, and as Dan Ariely points out they can be dangerous.
But economies of scale alone don't quite answer the question, because we could always counter that by asking why "bread" or "chips" aren't listed on the menu for a nominal charge, even if they're unlimited. The broadcast hints at the answer by implying that people don't like to be "nickeled and dimed" which is part of it - but believe it or not, restaurants don't benefit from nickeling and diming either!
Market transactions create costs that aren't present when we leave them unpriced, kind of like friction. A common example (which is sometimes also applied to the related topic of property rights) is how to price fresh produce. In this case, it's quite possible that the restaurant and consumers both benefit from leaving bread or chips unpriced, since pricing them would involve greater mental cost, monitoring, and in the long run make the combined bill somewhat more expensive.
So, if the cost of producing a good is small relative to the transactions costs, then a price of zero is quite efficient. Further, if the side-benefits of supplying the good at a price of zero exceed the production costs, it will be in the firm's private self interest to supply that good at the efficient price of zero, and without such side-benefits (even if they do arise out of a "prisoner's dilemma"), the good may not be supplied at all! In fact, if the transactions costs of exchange exceed the production costs, it's possible that by economizing on transactions costs, free bread really is free.