"May" being the operative word
"Google Flu Trends may provide an early-warning system for outbreaks of influenza."
"May" being the operative word. It's not designed to do so. No-one can rely on it to do so.
Google Flu Trends is an experiment, not a service. It's just a wild guess based on circumstantial evidence. The WHO is still going to rely on actual health data.
To go from this to "people will die if we can't do it" is ludicrous.
OTOH, long term, there may be some public benefit from mining search data. If this is the case, the data should be available to all in a public anonymized database, not owned by Google as a side-effect of their advertising monopoly.
"Which raises a question: if the point of using web searches to track disease is that the data is instantly available, how does data that is more than six months old help, let alone make us all less likely to die?"
If you suspend disbelief about the whole idea of Google Flu Trends, this is easy to answer. Suppose Google has got some formula that translates search terms into flu pandemics, but it's not 100% perfect. By analyzing historical data against actual pandemics they can tune it. This probably requires going back more than 6 months.