Two causes appear possible - residual stress and hydrogen -since this appears to be a delayed process probably a combination of the two.

This is a common observation for steel balls if they are not properly heat treated. After the hardening the tempering process should be able to temper all the martensite and there should be low residual austenite. Any large presence of residual austenite will begin to transform (not necessarily in the mill even while stored in the yard) due to this there is a  volumetric expansion resulting in metal splinters flying all over. At night it presents a scary scenario.  

arunmrao;
More from a curiosity viewpoint, any reason that the rock crushers wouldn't use Hadfield manganese steel balls that would work harden in service and achieve the same results without the danager of fracturing or splintering?

I am sure Climax Moly and others will have tried this in their mines. Perhaps due to initial faster wear out in high manganese steels and overall a higher wear rate as compared to quench and tempered steels the latter is preferred.

While on the subject of crushing have you used a crusher hammer having cast insitu ceramic in manganese steel. On the working face of the hammer there are a number of cylidrical holes into which ceramic inserts are placed. Now this is claimed to give a better life.
Alternately there is also a mention of bimetallic hammers where the working surface is high chrome iron 15Cr grade and the rest is carbon steel.

I shall be very thankful if you can share your thoughts or experiences.