I am a big fan of Coursera.com, but I did not know about Lynda.com. Figuring out how to monetize these offerings seems to be one of their biggest common challenges. I wonder if Lynda will have trouble competing with Coursera because Coursera courses are free if you do not want a certificate of completion.
Pay-as-you-go models are interesting to me because for some things they work very well, and for some things they do not. For example, the prepaid cellphones are an example of when it works well, for all the reasons you discuss. In general, in my mind this is because the expense is relatively small, and the consequences of not being able to use your cellphone are typically limited. In contrast, an industry like health care is much less attractive for a pay-as-you-go model because the costs of health care without insurance can be debilitating. In fact, approximately 50% of bankruptcies in the U.S. are the result of health care costs. The point to me is that it is excellent when a pay-as-you-go model is mutually beneficial to customers and suppliers, but one must also be wary that such a model could be used to exploit customers.
Your point about food purchasing not being performed by the government was especially well taken for all the reasons you listed above. Nutrition is one of those determinants of health that has a multiplier effect when provided adequately. In another developing country, for example, a research study included two groups with a chronic disease. One group received medication to treat their condition, while the other group received medication and a food subsidy. The group that received the medication and food subsidy not only had better disease outcomes, but attended clinic visits more regularly as well.