The 8296 appears to use High = Black and Low = White. The 8032 (CPU board from the SuperPET - and hence the SuperPET monitor) appears to use High = White and Low = Black. This accounts for the inverse video we are seeing. Have you tried adjusting the screen brightness ?
The 8296 appears to have a completely different circuit for the video sub-system, even though it appears to work in an 'enhanced' mode. I don't, however, see any reason why the PETTESTER should not be driving it correctly.
A 'normal' PET has separate video RAM. The 8296 uses 'system' RAM and shares this with the CPU.
My PETTESTER has clearly worked out that the 'video RAM' = Main RAM (in the area $8000 to $87FF) is working OK (otherwise it wouldn't move on past the VDU RAM check screen. It has also checked out pages 0 and 1 of the RAM.
The next thing to do is to compare and contrast the 80 column display in my PETTESTER documentation to what you are obtaining (and observing) on the screen. In particular, look at the first 256 characters at the top of the screen. Are the text and graphic characters displayed correct to my documentation?
Next, look at the 'fixed text' that should be observable on the screen, the 4-digit checksums for each of the ROMS and the 2-digit countdown value. Are these correct or not. Is the 'rubbish' that appears half-way down the screen fixed or does it vary?
I suspect some problem from the CRTC address and data bus through the buffers to the main RAM to be at fault somewhere.
I would be looking around UC9 (6545 CRTC), UA9 (74LS257) and UB9 (74LS257) first.
Dave