[Girchuck]

I attended a lecture on the subject in summer of 2004 given by a CDC scientist.
He suggested that given what they know about the H5N1 virus and the way virii swap genes, the probability of mutation that will give this virus high human-human transmission rate is 60% over the next five years. Most probable mutation route would be to swap some genes with a regular flu virus that would give H5N1 the human flu transmission rates without significantly reducing lethality. This is possible not only in humans who are infected with both virii types, but also in pigs who can be infected with both virii types as well. In fact, pigs were the most probable source of 1918 virus.
Once the H5N1 acquires the the genes enabling human-human transmission, there will be at most two weeks to respond before pandemic affects half of the earth population. In two weeks, the pharmaceutical companies can produce and distribute at most a million doses of effective vaccine.
If the mutated virus does not acquire resistance to antiviral drug Tamiflu, the spread of the virus can be checked by administering the drug to everyone at high risk of exposure as a profylaxis measure. In this optimistic scenario, US casualites from the pandemic will be only 200,000 (which, compared to the regular flu death toll of ~30,000 is not that bad). If the virus manages to acquire resistance to anti-viral drug, the casualties will be certainly in the millions.