I commented that my money was on experimental error being the more probable cause of the interesting results.
John responded: And yet millions of people around the world are treated successfully with homeopathy (including myself). Yeah, I know about the wretched placebo effect, but consider this: there is a growing use of homeopathy in vetinary medicine, where it has a similar level of success to humans. Just how DOES the placebo effect work on cows and horses?
I think that's a separate, and much more interesting, question.
As someone trained in science, my first response to an apparent anomaly is "Consider the likely causes". Of course, 'likely' has a degree of prejudice to it, but scientific training does help one to be aware, at least, of one's prejudices.
Faced with two possibilities (not that I'd deny there are more): one, that something very peculiar and not yet explicable is happening to water; and two, that something went wrong with the design, the carrying-out, or the analysis of the experiment - and error from the blatant to the deeply subtle is all too familiar to anyone involved in experimentation - then I don't think it's unreasonable of me to start, at least, from the assumption that error is the more likely cause.
As to why homeopathy may seem to work in animals - and I'm using seem without prejudice - there are well known problems in experimental design that can lead to animals appearing to behave in quite unusual ways. See any discussion of 'Clever Hans', the horse who appeared to be able to count. The Hawthorn Effect, while it's unlikely to influence the animals themselves, may affect the people around the animal - and I'd be the last person to claim that animals don't respond to people. It's not hard to find further examples of the difficulties that beset experimenters, particularly when they're looking at what may be very small or subtle effects.
Experimental science is, in short, a minefield, and there are so very many ways to go wrong with confidence in experimentation that a great deal of humility is a prerequisite when determining that a set of results 'must' mean something.