The GMAC will NEVER release any information regarding this as it would ruin its R&D process for new question development and implementation.
That isn't really true - there's actually an official research report that explains how experimental questions are used on the test. It's just that even if you know how they're used, that doesn't help you at all - you still won't be able to tell during the test which questions count and which don't.
Experimental questions are organized in 'batches', and I think that's done so that the difficulty level of everyone's experimental questions is similar. Before you start your test, a random 'batch' of experimental questions is selected for you. In addition, a random set of question numbers is chosen, and that's where your experimental questions will show up. Since this is all done completely randomly before you start your test, you won't have any way of knowing when you will see an experimental question, and nothing you do during your test will affect where your experimental questions appear. Your best strategy is thus to just assume every question counts and try your best to answer each.
Nov 2011: After years of development, I am now making my advanced Quant books and high-level problem sets available for sale. Contact me at ianstewartgmat at gmail.com for details.
Private GMAT Tutor based in Toronto