Disassembled:
3rd Gear Circlip, Beveled edge was facing thrid gear, sharp edge towards 2nd, as it should be. Wear marks are on the 2nd gear side of the circlip, indicating axial force applied during 5-4 downshift:
4th Gear Circlip, Beveled edge was facing 4th gear, sharp edge towards 5th, as it should be. Wear marks are on the 4th gear side of the circlip, indicating an axial load applied during 3-4 upshift! This is a new data point, I'll be looking more into it. It means 3-4 upshifts can also cause loading across the two splined bushes that 3rd and 4th ride on, and into the 3rd gear circlip. The key though, is the 4th gear circlip stayed in place, so as it applied load, the 4th gear circlip would have limited the amount of force transfered to the 3rd gear circip when the splined bush slip away.
3rd gear splined bush, you can see on the face wear, how it's getting mashed against the circlip:
This is the 4th gear splined bush, by comparing you'll see less wear, indicating the second mode of axial loading through 4th into 3rd is not causing the same load that 3rd is applying to it's splined bush or circlip. This supports the primary failure cause being 4-5 downshifts.
I also need to add, the 3rd gear circlip has far far less tension than the 2nd or 4th gear clips. Clearly, it's been beat and stretched over and over and eventually slipped up, and over, the groove face it's supposed to ride in.
This all supports the idea of cutting and using larger, stronger, circlips to extend the service life of the clip and prevent gears mating in abnormal ways.