I didn’t see when the shaft broke, but I saw both halves afterwards, and in this case torque was not the factor, it was bending stress from an impact on the cantilevered wheel - on the fracture face you could clearly see where the break originated.
To better clarify, I believe this snap ring groove was just outside the frame, between the bearing and the wheel on a WCD setup. It is a 1/2" hex shaft, and all the grooves had E-Rings. At a glance, the ring grooves seemed quite deep - I didn’t measure, but the remaining diameter was about 3/8". After looking up E-ring groove tables, they weren’t far off on proper depth, as the groove diameter on a 1/2" shaft e-ring groove should be .396"
As Chris mentioned you have to consider what the stresses are in a shaft where you intend to put a retaining ring - in this case, it was in the middle of the shaft where both significant torsional loads (from driving the wheel) and shear loads (from the cantilevered WCD wheel) existed. As DampRobot mentioned, sharp corners are the worst for stress raisers. The larger the radius at the bottom of a step or groove the better, but we tend not to do this when machining grooves. Also, the smaller the groove depth, the better. I looked up stress concentration factors for these e-clip grooves, and its somewhere around Kt=8+ for bending, and Kts=4.5+ for torsional stresses, which are pretty severe and make it clear why the shaft broke here. (This means the maximum stress from bending forces due to the stress raiser will be ~8x what the average stress is in the normal section of shaft.)
An alternative to an e-clip is a standard retaining ring - they can be more difficult to work with without the proper pliers, but they require a slightly narrower and much shallower groove, leading to much lower stress raisers, in this case the stress concentration factors would be around Kt=3 for bending and Kts<2.5 for torsion (you may also note the thrust ratings for the retaining rings are much higher than comparable e-clips, despite the shallower grooves). Its also not quite standard to work with hex shaft (good luck finding a table for it), but I bet you could even cut shallow of their recommended groove depth for a 1/2" shaft, as the corners of the hex add effective depth for engaging the retaining ring.
Also a very important point that Chris tried to clarify: any type of retaining ring on the end of a shaft is basically a non-issue, as no torsional or bending stresses pass through that section of shaft, and the retaining ring only sees thrust loading along the axis, which is exactly what it is designed for. So in this case, they could have used standard retaining rings at the mid-points on the shaft for significantly lower stress concentrations, and still used deep e-clip grooves at the ends to make changing wheels easier. Alternatively, they could go to shaft collars or properly sized spacers between the wheel and bearing to avoid stress risers entirely (just don’t expect tremendous thrust load resistance from a shaft collar alone, and realize thrust loads with a spacer will be carried through to the retaining rings on the ends of the shaftl).
If you want to check stress concentration factors yourself, check out Figure A-15-16 and A-15-17 on pages 1031-1032.
BTW - Great job dealing with unexpected damage and playing through without getting discouraged. Its always good to hope for the best, but design and plan for the worst, and never give up - its always better to put as much effort into a quick fix with the time and resources available than to give up when something unfortunate happens.