The 486 won't read it either.
Weird thing is that I can get it to detect that the drive is connected and identify the model, but not read the data. I suspect there may be a issue with the cylinder/head/sector counts being off, as I have had similar issues with a Conner CP-3044 that was detected but the CHS had to be manually set. The online numbers didn't work, but trying all the different "Type" numbers in an old 286 eventually found the type that worked and I copied the CHS from there.
I'm going to switch gears for the Toshiba disk and try laplink instead. For Windows 3.1 that should do what I need for now. I'll circle back with the oscilloscope later.
Interestingly, the CP-3044 works in my TI Travelmate LT286-45, but plugging in a CF-IDE or SD-IDE adapter into the the LT286's IDE bus (even with no SD or CF card, and the adapter not powered) causes a black screen no-post. Something is being pulled high or low that should not be. A future project will be to learn IDE standard and figure it out with a logic analyzer. The toshiba drive will get wrapped up into that. I'm past "trying things until it works" - now I want to understand why it's not working at the logic level.
Thanks to all that offered suggestions! Will report back with anything I find.