In answer to your questions
1) Did you do something specific the previous time when the issue was resolved? - NO but then the virtual machines have run slowly and as tests were not robust we reduced priority of monitoring the machines to see if it was still happening so my previous comment may not have been true in that it had gone away
2) If I understand you correctly, you again see a message "There was a problem starting ArtOfTest.Connector_64.dll The specified module could not be found" on your Windows 8 machine? Yes, i am getting this as a pop up.
3) Can you try to reproduce this issue with our latest internal build 2013.1.1120. You can download it from here. Have you actually done anything that will do anything different or are you just hoping? If you have i will try.
4) Could you please send us the full log file and a test file where you reproduce this issue. Please find attached the log file where the first test does the API call and hence causes the second test to fail with the error. This was reproduced using VS2012 IDE to execute the tests on a Win8 x64
Please note I have found out the following which is the likely cause...
I have been able to nail it down a little as to what the cause is but I am afraid that I will not be able to give you our application or test code as it all originates from tests that use one of our APIs of another of our applications which is sensitive. This API feeds the application under test with the data to verify the what is displayed is as expected. Let me however try and explain how it works in the hope you might figure out why all subsequent tests fail because Telerik Test Framework cannot connect to the application under test for the remaining tests in the test run.
We initialise the API in our test code which, as I am told by the developers, does the following when initiated within the invoking process:
Starts an internal time used for triggering memory flushing – This is never closed down
Initialises static factories used for dependency injection – These are never closed down
Initialises the service locator for unit tests – This is never closed down
Initialises the session manager class – This is never closed down
I understand these cannot currently change but what it suggests to me that when we use MTM 2012 to run our tests, something in Telerik (we use the same API in CodedUI using MTM 2012 without issue) is still clinging on to the API that was initiated in the previous test and hence why in the following test, when Telerik tries to launch and connect to the application under test which doesn't necessarily use the API, we get the issue we are seeing as telerik framework has not released disposed what's required to enable the next test to initiate.