A lot of time has passed, but it is good to complete information.

That link from Zozo follows to other link to a picture of three defective Philipines Z80 chips. On that thread I told about a lot of problems found while expanding a Microteam expansion card.

Later, I took out that Z80 to do a turbo hack on my computer. This is the back of the removed processor:

[ Guests cannot view attachments ]

My other EP was also in time modified to turbo, but it had a SGS Z80 clone installed.

Yes I know about ZX-M1 problem. But on the Enterprise the Dave already use the M1 for set memory waits, then immediately can notice if it is not working.
And on Spectrum M1 used for a "overwrite ROM" trick, executed address decoded in ROM, and the expansion page in their ROM, for modifi the original ROM program. This is the only way expand the closed system of Spectrum.
This trick don't needed on EP, because the EXOS have a nice expansion possibilities.

The bad Z80s what I noted: can't write to memory! Can read memory, can execute program, can read IO, can write IO... just memory write don't work... some are total, some are unreliable. All are from machine which is worked, stored for a few decades, and don't start now.
The problematic ones what Gflorez mentioned have a problem with memory expansions, especially with modern Flash/SRAM chips.
I think will be possible if retesting these chips after 5-10 years, then there is a completly no memory write problem...
It is a theory: plastick package of chips sometime not fully hermetic and the chip parts can corrode over the decades. Some lower quality production serie will make series of similar faults in a similar chips.
Similar example the Newtronics dead head problems in Commodore 1541 drives.

Very interesting!
Just for curiosity: what happen with two drives, COPY from one full disk to empty, formated disk?

I digged out two Korean VL chip:
This one, with 1988 w13 date code, works normaly.
[ Guests cannot view attachments ] [ Guests cannot view attachments ]

Another one from my bad chips box, date code 1987 w02. Looks same with Pear chip!
[ Guests cannot view attachments ] [ Guests cannot view attachments ]
And probably same fault?
I also made some videos (head step too quiet, turn up volume )
EXDOS FORMAT
The head many times go out to Track 0. Then overwrite previous tracks, not a surprise the Sector not found at the end...
Probably sometimes Restore command executed instead the Step In?

EPDOS formating. Total crazy, the head most of time restored to tr0...

FAFO formating. Works normally I don't know yet which are help: using my own WD177x routines instead of EXDOS DISKIO? Or just the reverse direction formating, where Step Out used instead the Step In?

Next test is a copy:
COPY disk to RAMDISK, with a good chip. Normal head movement, go to tr0 only for a next FAT or root DIR sector or file entry.

Same COPY with the bad chip. There is some extra head movement. The EXDOS retry mechanism handled it, no error, just about 15-16 secs longer the operation than the normal.

I also tried EPDOS copy, it is generate many Sector not found error, and press Retry to continue. (No auto retry mechanism) I think this is why I putted this chip to the "bad chips box" many years ago.

Also tested it today and worked normally. Tomorrow i can make few hours more testing.

Tested it for about two hours. Dozen disk formated, also including 10 and 11 sectors format, both with EXDOS and EPDOS. Disks copied with COPY, EPDOS Copy and Diskcopy. Working flawlessly.

FAFO formating. Works normally I don't know yet which are help: using my own WD177x routines instead of EXDOS DISKIO? Or just the reverse direction formating, where Step Out used instead the Step In?

I asked Bruce for make a EXDOS 3 test version with a reverse direction FORMAT.
There is the results, with the bad VL chip.
After few track the head restored to Track 0. Because the disk are previously formated, then the verify phase also can run. It is many time reseting the head, but the diskio auto retry handle the problems.

Summary: the problem don't depend from which Step (In or Out) used. But still there is the big question: how possible the FAFO are working with these bad chip?

The defective VL works properly in the Timex FDD3000 disk drive.

Least one place where it is working

In the original, WD1770 was used there.

EXDOS originaly also use 1770.

Maybe your FAFO works a little slower than the original FORMAT ?

No, it is FAst FOrmat
But very possible some timing are different. I wrote my 177x routines from a Enterpress article. Now I know it is a summary of WD datasheet. It is talked abaout command codes, but don't too much about timing. Then I set these by practical experience. For example the article don't talked about need to wait after write command for can read status. It is a surprise for me Then I putted some PUSH/POP for a waits. Is possible some better codes, which is used the original datasheet, have a shorter waits because use the exact timings.