Hi LucT, you are doing some fine detective work - however, I thought I should join in with some clarification...
I've taken some macro shots of my screen - apologies for any poor quality - due to compression, not the original image.
First two shots are the top left of the screen with default desktop (Aqua blue) with an image and some text on it. The second shot is a crop saved at higher quality.
Third and fourth shots are of the top right of the screen where a small screen capture of the top left of the screen is open at 100% size in a Preview window - so as to get an exact image comparison. Again, the latter shot is a crop saved at higher quality.
As I hope you can see, the descriptions by LucT below (about misaligned pixels) only applies to the leftmost part of the screen. And only in the blues and shadowed yellows/whites. It is my belief that the pixels or subpixels are not infact misaligned. The apparent misalignment observed, I believe, is the manifestation of the inversion scheme used by this PB display. That is: Line-paired RGB subpixel dot-inversion - see:
http://www.techmind.org/lcd/dotinvrgb2l.html
Notice how this test pattern matches that observed on the PB screen photo, and how it flickers most at the left hand side of the screen - precisely where the odd pixel pattern is observed in the screen photos. In case it is relevant, the shot was taken at 1/80th of a second. I don't know how frequent the inversion-scheme switches. (Does it match the refresh rate?)
Notice that in my test shots of the top right, the pixels are uniform in the composition of R G and B, but actually start to get bad toward the very right of the screen. The very right hand side of my screen also shows some slight flicker with the techmind test pattern.
Just to clarify: flickering is expected with one of the techmind patterns - it allows you to visually determine the inversion scheme type used by the display. The website suggests that if really bad it may also indicate poor display timing though - Is the PB's flickering excessive? - I'm not really sure...
Anyway. I'm not sure that the inversion scheme is the cause of the horizontal banding in alternate pixel rows since the banding is present in all parts of the display, not just at the top left, or extreme right.. but it certainly may be responsible for the wavy interference banding observed in (amongst other colours) the crankycat orange, which IS worse in the top left of the screen.
Keep at it. Now New Year is over, I'm about to start hassling Apple again (I've been out of the country for 2weeks+). Let's see what the new year brings
p.s. For any doubters out there - I hope these shots of the right of the screen are clear enough for people to see the nature of the horizontal lines that are being discussed: Each alternate row of pixels is slightly lighter/darker than the previous row. Nothing to do with the black outlines of each pixel.