I have seen this problem in my camera and mentionned it in another post to see if I was the only one. What I have noticed is that it doesn't show up in all photos, but seems more common in the upper ISO range and perhaps when I under expose. It might have something to do with the bayer pattern algorithm (wow, I managed to spell that, and no, I'm not an engineer). I noticed it in both my Jpegs and raw files so am assuming that the problem might be between the sensor and the card, ie somehow the way the image is processed in camera. Before you all jump up and down NO, I haven't done extensive tests but it has been noticeble in about 5% of shots, but not often enough in the same circumstances for me to say A-HA! However, if others are also talking about it then it definately could be an issue, but not something you can get around by trying different bodies in order to sort out. A hot pixel yes, but a pattern like this seems more like a programming issue. Smaller pixels evidently do have their drawbacks.. I must say that after being initially very pleased with my 20D over my 10D, my enthusiasm has waned slightly. Mind you I still love the camera and it's speed and usability issues, but a "faulty" sensor, whatever the cause, is a right pain in the rear end. I am hoping that another firmware update is on the cards that can sort this problem out.
If any others have some more insight into this issue I for one would really like to hear what you have to say.