Hi Bluesky,
No, but I will offer one possibility. Rick Alther says that the program checkpoints each time it successfully completes a refinement step (my terminology). HPF2 starts out with a crudely folded protein (HPF1 accuracy) and then tries to refine it using an all-atom relaxation model. I do not know how it decides how many iterations of refinement to try. But the point is -- a refinement attempt can fail. My idea (probably wrong) is that it shoves atoms into new positions according to force vectors until it discovers that it has shoved two atoms together. Failure! Superposition of atoms is not allowed. So it goes back to the last iteration and tries again starting with a different atom. So this sort of failure can lead to many attempts before it completes a new iteration and checkpoints. (Or possibly times out). So the time taken for each step is highly variable.

This is just a guess. But let it crunch for a while. I feel certain that there is some sort of time limit in the program code.

Lawrence