So it was quite clear last time that something had gone wrong with the Y-axis. It really was moving back and forth a full 0.1mm with a distinct click to it. So I opened up the machine again and pulled out the Y axis ball screw (Getting rather good at that by now :P).
The nice thing is that I realized I could do this without affecting any of the calibration of the machine. Just slide the X-axis over a bit to expose the Y-axis ball screw fastener bolts. Once done, reattach the Y-axis, slide back the X-axis and atach the X-motor to the bed. And done!
The issue causing the backlash seems to be that the spring steel pin wasn’t being springy enough and the screw bolt was not attached to the bearing fastener. What to do?
Fortunately, I found some solid metal pins of the right dimensions. With 8 layers of aluminium foil forming a shim, and some bearing glue, and the beasing fastener was rock solid again. Reshim the bearing house with an equal 8 layers of foil and put all of the pieces back together again.
Now the Terco has a backlash of 0.01 in Y and 0.02 in X. Unfortunately, the entire chassis is a bit springy, so the total flex adds up to something like 0.05mm. But at least we can now confidently say that the Terco performs at less than 0.1mm error in all scenarios.
Finally, the inaugeration of milling with the 4th axis. It worked like a charm! At least until I ran the end mill right into the chuck and ruined the mill and made a big mark in the chuck. Lesson learned: When doing 4-axis milling, model your chunk into your CAD model so you can verify that you are not cutting into it.
Also, Delrin is not as solid as I had assumed (Obvious in retrospect) so the vibrations were completely unreasonable if this had been anything but a test cut part.