CBM 8250LP update (hints welcome)

Hi all,
just a little update on the 8250LP drive I started some weeks ago.
Basically, when I started, it seemed like the 6502 "FDC" would not
execute any meaningful code. I found it had a dead PHI1 output, but
would work fine into a PET. Also I've verified that PHI1 isn't
connected on both the 6502s on these drives. In the end, by swapping
the two 6502, the 8250LP started to have some life.
It would however not read the disks (2) formatted with another 8250
(non LP) that I received together with the unit to be serviced (I
don't have any 100tpi drive myself).
I've changed the capacitors on the Panasonic mechanics and verified
that the nearby traces weren't damaged, I've removed the darkned
solder mask around some leaking caps and re-tinned the traces. I've
then verified and slightly tuned the spindle speed on both drives
using the index sensor that I wired to 5V with a pullup resistor and
ground, the sensors on both drives are not used by the Commodore
digital board but are wired to pins 21 and 22 of each drive's
connector. I've found both drives to have 201ms interval between index
pulses, so I slightly tuned both to have 200ms nominal pulse interval.
I believe -/+ 3ms is ok however.
I've then tried formatting new floppies and drive 1 could complete the
format on just about any floppy I tried (more than 20 of them, never
failed one). However drive 0 would never finish formatting even on
de-gaussed media.
I've identified the problem of drive 0 to be a marginal or defective
r/w coil on head 1, see these short videos:
https://www.youtube.com/watch?v=_IWwMZAJjcI
https://www.youtube.com/watch?v=BsykFp8qZw4
https://www.youtube.com/watch?v=fn6r5pwQEqo

The first one is read signal on drive 1, the format routine formats
first the lower side (head 0), then the upper side (head 1) then step
to the next track, You see there's no difference between the read
signal on head0/head1 on the first video, this is drive 1. The
interval when the signal disappears is a write phase, the signal is
too big to be seen with the vertical scale I was using.
Second video is drive 0 attempting to format, there's a noticeable
difference between head 0 (normal) signal and head 1. Of course, as
the drive progresses to inner traces, the radial speed of the media
decreases and the signal gets eventually too low.
Third video uses a lower vertical scale on the scope so also the write
waveform is showed. The write waveform is also slightly different when
it's switched to head 1.
Of course I've tried to swap the drives to exclude head select diode
issues on the analog board.
Out of desperation I've also tried to see if I could re-align the
upper head on drive 0. This isn't something to be attempted without a
reference disk and the right programs on 100 tpi drives, but after a
few hours tweaking the head position (I've got a macro picture of the
mounting position before starting), I haven't at least made it worse.
Drive 0 is still able to read all tracks on a disk formatted by drive
1, albeit with some occasional head bumps on media having a not
brilliant S/N (and only when reading tracks on head 1).

Now, the really puzzling part is: why I can't read the disks formatted
on the other 8250? All I get is 20 READ ERROR 39 0
Things that I've checked:
DS0/1 start at 1/1 then change at tracks 40, 54, 65
The GCR rom has the correct dump.
I've written a program that leaves the motor on, and can step in or
out by pressing a key on the PET, it seems the disks formatted on the
other 8250 have lower signal (as if alignment isn't really spot-on)
but the analog signal isn't too bad to justify a read error anyway.
Sync is clearly recognized or I would get a 21 and not a 20.
Is there something else I'm missing? The likely explanation would be
that two drives in the 8250LP have exact track alignment and two
drives in an 8250 also have exact alignment between them but one pair
is badly out of alignment?
This seems really unlikely to me.
Any hint is welcome.
Frank IZ8DWF