On my fast system (and infrequently on some buildbots), it takes a long time (potentially forever) to kill while(true){} spinlocks sent to the mongod server. This seems to especially occur when using v8, seems as if the kill operation just is not able to progress. Adding a sleep(1) statement to the while() loop fixes the problem.

While currently the problem only shows up in tests, potentially an accidentally buggy $where query may be unable to be terminated on a user's system while hogging the cpu.