Hi, Mike
It seems that I am currently facing the same issue in my work, exactly as you described.
During the implicit coupling between the fluid and particles (OpenFOAM DPMFoam coupled with CalculiX), when a coupling sub-iteration fails to converge and a rollback is triggered, the fluid field variables are correctly restored from the checkpoint. However, the variables stored within the KinematicCloud are not restored. As a result, during the next coupling iteration, the particles continue their motion from the already-advanced state, effectively evolving into the next time step again.
As a temporary workaround, I modified the DPMFoam solver so that particle tracking is executed only during the first coupling iteration of each time step. In practice, this approach works reasonably well and avoids the inconsistency. However, it requires modifications to the OpenFOAM solver itself, which is obviously not an ideal long-term solution.
It is actually quite reassuring to see that someone else has encountered and investigated the same problem. I was beginning to suspect that I might be the only one working on this particular issue.
Xiangxiang