I can only give you my personal experience about your question. Through working with ABAQUS, I was recommended to chose the increment time as higher as I can; because the higher time allow the software to examine all the possible cases which resulted to the best answer. I'm afraid if any one could give you the exact time incrementation. On the other word, considering the increment time strongly depends on your problem and the mesh sizes you selected during the solution trend. in general, for densely packed particles, you need to adjust mush higher time period than usual.
The main issue with the discrete elements is that they are rigid and a fixed time incrementation is required.
Using a high time increment for discrete elements may result in neglecting the contact stiffness. So it requires a compromise between numerical stability and analysis cost.
Unfortunately, there is not sufficient literature on this issue.
If you find any, it is appreciated if you share it.