A dice fiend myself, I would be interested in where one can get the d24, d5, d2s, and the skull dice.
As for the d7/d9: it is theoretically possible to make a fair die of any number of sides using that method. I have seen the d9's and they didn't look very well manufactured. On the other hand, the d7's are made by Gamescience IIRC, which is usually pretty good about it's manufacturing.
Back to the d5 the canadian prof tested, I'd be very interested if you have a reference on that. If he's using the usually chi-squared test for goodness of fit for discrete distributions, that's a very low power test. That is, it can tell you a fair die is fair pretty well, but it's not so good at saying an unfair die is unfair. At least not without a LOT of replications.