I found this reference: https://en.m.wikipedia.org/wiki/Commodore_1571
Something about the worst case data pattern for transition density.
Well that's interesting. I have 2 and half 1571s and I never looked that deeply into them. The "half" 1571 is the one built in to my 128D. They're very reliable drives IME so I never looked that deeply into them.
Sooooo I've been helping someone troubleshoot a SFD-1001 drive which packs 1MB on a DSDD disk, but these drives have trouble formatting the higher tracks where I guess the transitions are packed tighter.
My working hypothesis is that head wear changes the gap and therefore the frequency at higher tracks.

There is almost no information about floppy heads out there. So I've been searching for whatever ICs were used in floppy discs.