Treat unit and system tests differently
I've worked in a number of scenarios where we had a subset of tests which would be better classed as "system tests". I.e. they covered a large amount of system functionality and took a considerably amount of time to run (sometimes minutes). Running these as conventional unit tests under Ncrunch, triggered by every editor change is overkill but OTOH the Ncrunch UI is really nice and its ability to automatically run things is very helpful. It would be useful if these tests could be treated in a different way to normal units tests, e.g. only run on a scheduled basis (subject to editor changes) at a low priority with a larger timeout than that used for unit tests.