This has been tested within Juniper on a large-scale multi-threaded app. I also plan to run some simple pathological threaded apps (constantly starting/stopping threads, several threads with long-running tight loops) which I've used for basic PMC sanity-testing for previous changes.