NCrunch spirals out of control

At least four developers on my team plus myself have been struggling with NCrunch: After running tests normally for about five minutes, it will to to a state where it slams the CPU (pegging as many build threads as I've allowed). When that happens, I can shut down Visual Studio (2012) and about one minute later NCrunch will die. Then my machine will become responsive again. It did not do this the first week that I used the product. Now it does it very consistently. How can I debug what is going on? When NCrunch pegs the processor, it take about a minute for the Task Manager to start up, another minute each time I try to scroll in the Task Manager, another minute to get the context menu up to kill a process -- it's very painful.

Hi, thanks for getting in touch with this problem.

Can you describe the nature of the tests you are executing with NCrunch? Do you have any integration tests that interact heavily with the file system or processes outside the test domain? (i.e. a database).

If you can, I recommend trying to get a capture of the tests that are running when the machine starts to slow down or lock up. Working with the processing queue open and the filters set to show only the executing tests should help with this. My first theory is that you have at least one test in your solution that is managing to overstep NCrunch's processing constraints. If we can discover which test it is, we can then identify where things are going wrong.

It stalls during project build. It appears to only happen when you restrict NCrunch to specific CPU cores. Try this configuration for a few days:
NCrunch CPUS: 0,1
Visual Studio CPUS: 2,3,4,5,6,7,8
Threads: 4

Is there a specific project that it seems to stall on while building? Have you noticed this problem with any other projects/solutions?

It does not stall on any specific build or test. It may be that it always stalls compiling, though.