I am using the latest version for .NET 4.0 with NUnit 2.5.5 (using the /framework=4.0.30319 arg for NUnit). On small projects (i.e. simple tests that execute quickly), it seems to work fine. On a larger project that I have ported over, the command line sometimes fails. It seems that the code that waits for the external process to finish (in my case, nunit) attempts to generate the report too early (sometimes causing errors with invalid handles and sometimes generating an empty report).
I have done a bit of digging around in the source code and have managed to get it working consistently by adding a slight delay after the external process is complete (but before the statistics are gathered).
In Connector.cs - around line 76:
This is an incredibly nasty hack - but it's working for me at the moment. I'm not a profiling expert but is it possible to respond to an event raised by the CLR profiler to know when the process is complete rather than polling the process and waiting for a return code? (it may do this already)