A while back I read a piece by, I think, Joel Spolsky on the "design" of the XLS format. It was mostly a dump of Excel's internal memory structures to disk, with some optimizations to speed up saving and loading on the slow disks of the time. He seemed quite satisfied with how well the software performed but all I was thinking is "any other program trying to deal with those files is screwed." And "any other program" includes later versions of Excel.

It'd be nice to have some documentation of the internals of old DOC and XLS formats, just for the sake of recovering old files for archival purposes, but it's likely that Microsoft never bothered documenting them, even internally.