We execute one test definition (each instance has an unique test identifier generated automatically , for example 12345.
This test is then executed on many operating systems simultanious (currently each one box for each operating system). Each system can execute up to 5 tests in parallel.

How do we ensure that only one mail account is used at any given time?

The best I could come up with is below method, which I ask you to feedback on it it has any flaws?

Serverl_perl script for auto mail testing
1. Perl script connects thru ADO to a table on Database Server that is populated with working Exchange accounts in tblAccounts
2. Script retrieve an _unused_ account from an Accounts table for example Autotest_Account1 , then sets flag to _InUse_
3. Cleans that's accounts mailbox using Perl CDO or MAPI or similar
4. STRS test executes
5. Perl ADO sets tblAccounts - records Autotest_Account1 flag to _unused_

This makes sure each test always gets an unique account