making tests more reliable - detect duplicate manager processes?

[bertschinger]

sometimes if a manager process is not killed at the end of a test run, that manager process interferes with later test runs and causes them to fail.

could there be some way to detect this and alert on it so that when "flakey" tests are occurring for this reason, the cause is obvious?