Date: 27 Nov 97 07:14:26 +0100
It is indeed a 1542A, which has worked flawlessly for years in dos.
Will get your tools, and get back.

the disk is the 4'th drive, scsi-id 5.

So now I have tried testblock, and it shows that 0xff are read at 4K-boundaries:
0x2001, 0x3001, 0xc001, 0xd001.

What do I do now? Short of changing the controller?
Is read-after-write-verification possible? (I won't be doing much writing).
Or shorter blocks? 

My 1542A indeed often writes 0xff at 0x2001,0x3001 etc.

Have I just confirmed I'm in the same boat as you, or does a solution exist, short of getting another type of adaptor?