On Windows, apparently PIDs can be recycled in such a way that they can reappear shortly after the first process with a PID is terminated. (See http://stackoverflow.com/questions/26301382/does-windows-7-recycle-process-id-pid-numbers/31257979#31257979 and links therein.)

RespawnHttpTestCase is assuming that this won't happen, since when checking for reload completion it ignores any PID that belonged to one of the servers that's expected to be restarted. I'm seeing intermittent failures that seem to be due to this expectation being incorrect, as the logging for the failed tests clearly shows that the servers are restarted.

For example:

https://ci.wildfly.org/viewLog.html?buildId=45601&tab=buildResultsDiv&buildTypeId=WildFlyCore_PullRequestWindows