Agent Dimension in Call Center DW

My question is for anyone that has experience implementing an Agent dimension in a DW that supports a call center.  

Background

Our Agents generate activity in source operational systems including call management, workforce management, evaluation and order management to name just a few.  These logons vary from system to system (jasmith in one system and smithja12 in another).  There is no central repository for Agent logons and logons can be recycled as the call center staff experiences a lot of turnover.

The greatest challenge we face is that there is no agent natural key such as SSN or employee ID that can be used to consistently identify an Agent.  The workforce management system (Aspect Workforce Management) does generate what appears to be a unique and durable surrogate key for each Agent.  This package also allows agent level user defined fields that could be used to store logons for each Agent.

Question
My somewhat vague question is…. Has anyone implemented or worked with an Agent dimension under similar circumstances (no durable Agent natural key, multiple dissimilar logons, Aspect Workforce Management) and could you please provide some high level direction on how to best model a solution?

Thanks for your help!

Not agents specifically, but having different natural keys coming from different sources for the same entity is not uncommon.

The most important thing is to identify dimension rows using their true natural key.  If Jane Doe is in three systems and has three different natural keys, there should be three dimension rows.  Facts from a particular source would be linked using the natural key from that source.  This creates a stable link between transactional data and the agent as identified by the system where the transaction occurred.

To 'consolidate' transactions for Jane, you populate the dimension rows with the same attribute values.

This is a similar issue to the one discussed here: http://forum.kimballgroup.com/t3254-handling-customers-merged-via-mdm