In our company we have a callmanager 6.1 cluster. I needed the server for something else and I did a powerdown of the subscriber. The subscriber was the primary callmanager for the phone's/gateways and CTI ports.
The subscriber drive I put on my desk and used a new drive to make a LAB test.
The system was running on publisher for 1 month, and during this time I made several changes like speed dials/BLF, changes in directory numbers, hunt group configuration etc.
So we can be sure that data on pub and sub are not the same anymore.
When I switched on the subscriber, very weird things were happening, suddenly people were logged in or out (Extension Mobility) some BLF and speedials dissapeared, in short is was messy. I solved it by making the pub the primary callmanager for the cluster and after a reset all became stable. I did a cluster reset and when the database was synchronized, I changed callmanager order again.
So it seems to me that you should never do this, and if a subscriber is defect, but not the drive of the sub, but powersupply or motherboard, you should not simply switch on the sub and think that everything will be OK. Doesn't callmanager have a sort mechanism to prevent this?