Hi David,
The header (size and destination) is missing on the problem board, which could be the reason that the board is not starting up.
The AM3352 checks four different locations in flash for the presence of a header, at addresses 0, 128 KB, 256 KB and 384 KB, so if there's a valid header (and XLDR image) in just one of these four locations, the board should still be able to boot up. So just to be sure you should examine the contents of the first sector in each of the other three areas as well.
Another thing I notice is the bad block marker, which appears to be 0 for the problem board (reverse logic - 0 indicates a bad block, 0xff a good block). I would suggest looking at the first sector in a number of blocks (including XLDR and Eboot), just to see if sector 0 being bad is just a coincidence of if there's a problem somewhere causing blocks to get marked as bad by mistake.
Why this happens? Difficult to say without more information.
Have these boards started up successfully in the past? If not, then I'd suspect an issue with the flash programming routines (have you modified these in any way?) Do you program XLDR into all four locations searched by the CPU, or just the first one?
If the boards did start up initially and then developed this fault, well, I would suspect the WinCE FMD / flash driver, which should be the only one writing to flash once the board is up and running. In particular you need to ensure that all flash blocks not intended to be used by the CE flash driver are marked as reserved.
You mention that 10 units are showing this error, but what is the error rate - how many have you produced in total up to this point?