In SOFT=2 contact, with the sliding option (SBOPT=4 or 5), both neighbor node and neighbor segment normal vectors are used to judge which segment is being penetrated and which is doing the penetrating. It can be hard to judge this without neighbor segment data, and so the sliding option reduces that chance of tangential force spikes or other bad behavior during sliding. One of the reasons to avoid this is that there is added MPP message passing to get the neighbor segment data set up correctly. This is done in two steps. In the first step, data is passed if needed so that each segment has a complete set of its own neighbor data. In the second step, the neighbor data is sent along with the segments that are sent for the contact calculation. Message passing takes time and can cause processes to wait for others that are slower to send their data. The second issue is that there when the sliding option overrules the initial judgment of penetrating and penetrated segments, it creates a possible discontinuity in the force magnitude. The default judgement is based on the minimum penetrated segment is not susceptible to force spikes and shooting nodes, but when this judgment is overruled by the sliding option, the force may spike. To prevent this, the stiffness may be reduced, but reducing the stiffness will increase the chance of excessive penetration. The bottom line is that the logic is not smart enough and so it is not fully trusted that does not increase the chance of penetration. Those are the two reasons for not recommending the sliding option unless it is really needed.