I have seen this on belt drives that were over-constrained (hard-coupled) between the belt and the pulley. And it was designed by Parker Motion. After running these stages through reliability testing we saw more than a few failures and Nema23 shafts shearing in half because of cyclical loading fatigue. The pulley and shafts were supported on both ends with a large bearing(s) and the shaft mated to the motor shaft via a hard coupler - the bearings supporting the belt tension were on the shaft with the two bearings, not even on the stepper motor.
It appears that your motor shaft failed in the same way as mentioned above. No clear defect or void can be seen that would cause premature failures from stress concentrations etc. The failure does appear to be close to a feature that reduces or necks the shaft down to a smaller size right at a bearing constraint. Your pulley makes a belt contact at the furthest point away from this bearing constraint along the length of the shaft - in other words, your belt is creating a large moment (due to this distance, and tension) around a feature that could be more subject to stress concentrations.
It is always bad practice to hard-couple or put belt tension on your drive engine shaft. One can imagine a process that perfectly controls concentrities between mating shafts that would mate a motor to some other rotating component perfectly, but this won't readily apply to a good number of cases; processes are difficult to control while removing the failure mode all together by use of a flex coupling eliminates your problem.
A larger diameter shaft may solve your problem, or just push off the failure mode to much later. You had a 1/4" diameter shaft fail within 10 hours, do you know how much time an extra .125" (3/8" shaft) diameter will buy you in terms of lifespan given all the other conditions are equal? That's about 2.25X the cross-sectional area. You should be able to calculate the life.