unlikely the card would report 00's if in fact it wasn't. usually a problem reading flash will show I/O error, not zero's.
exceptions to this is if you told the software to ignore IO errors, and in this case, I know DMDE can report Zero's.
the fact that good (but unwanted) data is available leads me to think the format did in fact overwrite the wanted data.
I don't hold much hope for this case. hardware recovery I assume means chip-off and nands or monolith read at a nand level. I guess if the FAT was showing zero data as in a quick format, there may be data to be retrieved. but video is more problematic to recover. I think the only thing left to do in this case is to read the nands in a chip-off recovery and look at whats there.
would anyone know (without testing) whether data would typically be data be written across the entire SD card randomly, or sequentially?
The data would appear to look sequential when looking at it through utils in windows, as will any FAT filesystem. not talking about clusters being allocated here I am talking about the FS image.
in fact, data itself at a base level is stored in whatever algo's the controller uses, and could involve rotating pages in blocks of data a certain size, pairing blocks of data, XORring the data with a key, or all of the above and more. It would likely look like gibberish in a hex editor