The motherboard is not identical to the link. I think it is this one:
http://www.uktsupport.co.uk/apricot/mb/aprtrent.htm - the physical layout looks identical, as do the BIOS screens.
The motherboard says Trent Motherboard, and Apricot Computers Limited (c) 1996, It is PNo. 13299631 (listed on this site, but with no useful details.
http://insight.actapricot.org/insight/products/desktops/ms/mstntbrs.htm)
Anyway, it is similar, both in terms of vintage, and in having a riser card for the card bus. There is a network card, plus a proprietary full length slot with Xilinx chips controlling the XYZ axes and other I/O of the CNC. Oddly, there is a little jumper cable from an on-board serial port to the main controller card.
There is another version (the original?) here:
http://insight.actapricot.org/insight/products/desktops/ms/mstntlay.htm - this seems to have a bit more information on hard disks, network cards, etc, but I am not sure what exactly to make of it.
I am pretty sure the error was "1780" (Fixed disk 0 failure) but can't swear to that (may have been "1762"). From memory the disks did show up legitimately (correct size, perhaps even model number) on the POST/BIOS though (but it has been a while). I have seen some other online hints that it may be a cylinder limit in the BIOS.
Generally I am not that hopeful that imaging to a completely new "vintage" PC (and installing the custom cards) will work. It seems to be from an era where the "drivers" are specifically expecting a certain type of chipset. ie. Apparently the network card has to be either 3COM chipset (which I have) or something else, which I forget.
My ultra-long-term plan is to completely re-automate the machine using Mach3, LinuxCNC or some other similar CNC controller solution. But that is a huge job, and will have months of downtime, so I would rather stay with the 90's era controller, as I have not found it too limiting most of the time. However if the PC itself or controller board die, then that is my plan.
Back to the Hard Disk geometry, it sounds like you have understood my confusion. But I have read various online resources about OS's going direct to the IDE controller (skipping the BIOS?), among other things being able to overcome various size limits. I started making wild theories that a real-time OS might set up a file system specifically optimised for low latency, or at least deterministic latency.
Yes, the HDD interface is ATA IDE.
Yes, the jumpers on the similar-yet-different HDD's were definitely set to Master on all the (3?) similar drives we tried, having cloned the image on to them. Sure, there would have been some blank space at the end, but the data that was read should have been identical I would have thought.