The old DEC VAXstation VSXXX-AA mouse designed by Hawley used a XY tracking arrangement of two roller wheels with their axes canted at about a 10 degree angle, one for vertical and the other for horizontal. This not very popular design was pioneered by the original Xerox mouse (featured in "The Demo") and also used by Honeywell for a PC version in the late 1990s.
The original roller wheels of the VSXXX-AA were made from a highly perishable hard rubbery plastic material that could only be described as "craptastic". On mine the rollers looked like a wheel of Camembert cheese (see photos), with a similar white layer of mold. These crumbled at the slightest touch, leaving the remains of the roller on one, and nothing at all on the other.
This small print job is a workalike replacement. At first I contemplated machining new roller wheels on my little Unimat lathe from a 5mm sheet of polyurethane scrap, then thought up this idea. Simply put, it is a button-like base with a seat and curved retaining lip for a small neoprene O-ring. The O-ring projects 20% of its thickness above the surface of the wheel face to make contact with the desktop. I added two holes for easier wheel removal using a pair of fine needlenose pliers or bent paperclip.
You will need to disassemble the mouse to enable proper press-fitting. To do this, use a flat bladed screwdriver to pop the two side catches and rear catch in the mouse body then ease it off. The roller wheels and the optical encoders are on small steel axle pins, held close by weak magnets to impose a light downwards pressure on the wheels when on the desktop.
I printed these in ABS for some degree of flexibility when press fitting onto the shaft. PLA would probably be fine but might not be as flexible.
Set the TOLERANCE value to 0.5 (half a mm), F5, render with F6 then save the STL and slice, and run a print. Even at 100% infill this only took me a few minutes. The print may fit but is likely to be a little loose on the shaft. Bump TOLERANCE down by 0.1mm to 0.4 and run another, test. Do not make them too hard to press fit or you risk damaging the encoder or the shaft housing - it is better to bump the TOLERANCE up and down a trifle and run another test!
The O-ring I used was from one of those cheap assortment kits found at "two dollar shops". If you need to, measure the O-ring you have and adjust the parameters as needed. I don't think it's too critical - the shaft fit is much more important to get right.
If the O-ring is too loose then you can secure it to the roller wheel with a tiny bit of superglue, so far I have found the 20% protrusion amount sufficient to hold it on without gluing.
Printed from ABS to give some flexibility when press-fitting to encoder shaft.
Extruder temp 235C, bed temp 90C, print speed 40mm/s, 100% infill.