Thanks for the confirmation.
I appear to be seeing the same problem with CentOS 6 / CentOS 6 combinations:
| 9,931 | srv-c601-backup | 2015-05-01 23:32:01 | B | I | 90 | 12,990,648 | T |
| 9,940 | srv-c601-backup | 2015-05-02 23:08:27 | B | I | 3,877 | 50,585,887 | T |
| 9,949 | srv-c601-backup | 2015-05-03 23:08:56 | B | I | 101 | 13,051,119 | T |
| 9,958 | srv-c601-backup | 2015-05-04 23:09:04 | B | D | 87 | 13,025,155 | T |
So the relative age of the Director and FD does not appear to be a factor.
That's what I tried at first, as described in the paragraph you quoted; it left files out as described. Including all the available incrementals and differentials got me close enough to get the system back to functionality after a lot of "yum reinstall" to fill in missing system files. Fortunately the system was a relatively new one, and I keep a copy of /etc/ in Git as well.
The headline version of packages shipped with enterprise distributions is often pretty ancient, that's the tradeoff you pay for stability.
I take your point, though, that the Director/SD should not be older than the clients, so I need to fix that. Fortunately the (virtual) machine running the Director and Storage daemons is dedicated to that task, so it should be relatively easy to build a new CentOS 7 machine to get 5.2.13.
I don't think I actually have a version mismatch problem (as I'm seeing the same issue with matched versions), but there are all sorts of reasons this might make my problem go away: there may be a bug in the version of 5.0 shipped with RHEL/CentOS, or I may have a configuration problem. Either way, starting from scratch and transitioning clients over may help.
In the meanwhile, I'll probably just stop using differentials.
Thanks,