Just thought I'd pop in again and share my weekend's worth of work - a more scientific look at how scanning speeds, drives, media type affect scanning, and how repeatable scanning results can be.
There is a lot of data, but I've tried my best to present it well. I think there might be quite a bit worth discussing - if only just to say that it seems that the 4x scanning speed definitely has merits, as does 2x (although time is a problem at 2x). Scans of 6x show slight elevation of errors, and 8x quite significant elevations especially towards the outer edge.
The repeatability of scans surprised me - it seems that the scans on the whole do vary somewhat from scan to scan, such that the decimal places after the averages don't seem to be too significant.