We have had quite a few outages due to "concurrent service". My impression is that the IBM folks making the changes simply don't know what is safe and what is not. I don't know if the documentation is wrong, or simply too complex, or whether the people writing the documentation assume there are never any software problems, but it keeps happening. I would give IBM a few more years to work out all the bugs.
We test software changes in a test system before we put them into production, why not also test hardware changes? Many (most? all?) hardware changes are really microcode changes. And microcode really is software! Why should microcode be any less buggy than other software?
