For example, I do it by
gradle undeploy & gradle undeploy -Dport.offset=100 &
in a shell (just to reproduce the bug). I was able to reproduce the bug this way at least for 2.14. The real usage is a bit different (a more sophisticated shell script, but basically launching just two subprocesses in the same way, maybe with a bit larger time shift between the launches).
Until now, I was not able to reproduce the bug in 3.1 nightly build.