I've got a solution with DOZENS of projects and hundreds of tests. This solution is so large that I don't expect to get any kind of specific answer as to what's causing my problem, so I'm just looking for some insight: how does NCrunch determine when a test is 'impacted'.

I have found that the "Run impacted tests automatically" engine mode works fine many times. Other times, I've found that a change has somehow triggered dozens of tests should not have been impacted by the change. I can imagine that there are any number of configuration settings in any number of these projects that may be unnecessarily or incorrectly set that is triggering some sort of change detection, and I'd just like some input as to specific behaviors that I can look for and specific settings I can tweak that will cause only the barest subset of tests to be run under this engine mode.