I've not encountered this situation before, but the following checks may help:
Check your version of the QlikView Server and if there have been any bug fixs since your version that may be related to this issue.
Check your PGO files are in sync, specifically the CalData.pgo
The first is located in the Root folder as set in your QVS, such as C:\ProgramData\QlikTech\Documents
the second is located in the folder C:\ProgramData\QlikTech\QlikViewServer
You could also check in the QMC under Users, try searching for an example user such as Christoph Goeker to see if the user is returned twice.
They should only show once, so that might point to a problem with your configuration.
Other than that you might need to contact Qlik support to see what is going on.
In this view the Management Service will only fetch data for the select QVS.
To verify that the CalData.pgo also has the same information as you are seeing in the Management Console, you can add a setting to the QVS.
In Settings.ini, default location C:\ProgramData\QlikTech\QlikViewServer, add PgoAsXmlAlso=1 under [Settings 7]. You need to stop QVS while changing this.
This will make the QVS to also create a live copy of the pgo file in a nice readable XML format (located beside the pgo files).
Please try this out to figure out if the duplicates resides in the PGO file or within the QVS code.
Remember to remove this setting once you have completed this test.
Sorry for the late reply, as this is an older discussion and you still seem to suffer the same anomaly.
If you have two DSC's that service the same domains, you should cluster them. Otherwise both DSC's will respond to all ID queries and you'll get double assignments.
DSC's can be clustered by not defining them in System->Setup->Directory Services Connectors (green + icon), but by removing one of the unclustered DSC's, going one level down to the tree entry of the remaining DSC and clicking the green + icon in the General tab under "Cluster". You can change the name value to reflect that you have a cluster instead of a single DSC with a specific machine name.
The advantage of clustering DSCs (aside from not getting double CAL assignments anymore ) is that your domain queries will become resilient (one DSC can go down) and the load is distributed.
Note that you don't need a cluster license to do this.
[Edit] Corrected a lot of spelling errors. I seem to be getting worse at using keyboards :-(