How much money you got to spend? You can always ask for a Class 0 chip, but it will cost you. Producing CCD chips isn't cheap, nor easy. The technical tolerances to producing a Class 0 chip are extremely high. Most research facilitates will settle for a Class 1 or 2. The higher the number the more defects. Most camera manufacturers will tell you their cameras are within specifications for defects, and to be technically accurate they are. I know of an Apogee U42 back illuminated camera with hot pixels and a slight hot spot which is within specifications....and this is a US$45k camera.
It is rare that guiding software will guide on a hot pixel given the initial sub exposure identifies star centroids. Hot pixels don't have a centroid, but I could understand the software making a mistake if sampling is less than ideal. As you've suggested 2x2 bin can greatly assist in both sensitivity and sampling. In addition, dark subtraction of the guiding exposure will significantly boost the S/N to identify the ideal guide star.
You're going to have a tough battle claiming defects as a warranty issue. Chip characteristics change with time. There is plenty of info online about chip classifications such as standard grade, research grade along with the various classes. Here's a start on SBIG ST series -
http://www.sbig.com/FAQ/ST-7_8_9_10_...ass_of_CCD.htm
EDIT: Specifically to address your issue of losing a guide star, it doesn't hurt to script the reacquisition of a guide star between sub exposures or check out some of the MaximDL plug-ins. Multi-guide is pretty cool. Allows you to guide on more than one star simultaneously. At the end of the night however, clouds are clouds. Loss of a guide star is sometimes unavoidable.
