To understand what is going on, you have to remember that each task is split up into smaller chunks. When a chunk is complete, the agent saves the progress (a checkpoint file) and continues. If the agent is interrupted, then it has to go back to the checkpoint and continue from there.

Now, for estimating progress the agent assumes all the chunks are the same size. However, it is actually impossible to tell beforehand exactly how long it will take for a particular chunk, so it may take more or less time than the predicted period. So, when it gets to the end of a chunk and it's ahead of the predicted time, then your progress will jump forward. If, on the other hand, the chunk takes a lot longer than predicted, progress will appear to stick at the end of the predicted interval. It may even go backwards in rare cases. This is what you are seeing here.

The best thing to do is to leave your agent running uninterrupted (give it a day) until it has got over the tricky part. If that doesn't work, then report back and we'll look into it further.

Same thing for me. I run FAAH and it blocks at 33% and sometimes falls to 32%. Just before, It fell from 33% to 0% after a reboot. BTW, the best docking energy is approximately -25, which is unusually low.
Regards

Yes I know but the task does not finish and that's the big problem ! :-(

Thank you for explaining it to me. I understand it a lot better now, and I am releived that it is not a problem with my computer.

Left undisturbed it has moved on to 42%............will let her run and see how she plays out.