"Several dozen HN users will go to the website, see which variation they get, clear their cookies until they get the right one, then pay me $30... and then cackle madly" is not a real worry.
But, I do think publishing the A/B test does influence the outcome if they do overlap.
You are talking about the $30 that you could get but miss the point: you might get some extra money but in the process corrupt the validity of your A/B test and make a bad decision afterwards, based on that result. Well, not that it matters anyhow, as I'm pretty sure you have actually decided already to stop doing the Java app.
I would like to have somebody that knows more about sociology and statistical surveys explain if what I am saying makes sense or not.
I think you missed his: it's pretty unlikely that anyone on HN will do that.
Though as you say, the target market and HN crowd isn't exactly overlapping but there may be curious folks who can skew the results or the difference of performance between variations.
But, if people know they are being A/B tested, this allows them to make a decision about what they think they prefer.
Sometimes what people think and are willing to vote (say, by complaining they don't like their option, or by buying a product to support it) isn't what they would have actually picked.
I mean, I don't think HN readers are in the bingo cards market, but still, it feels like he might be influencing the results by blogging about this beforehand.