test 6, "allocate_as_barier_proc", failing intermittently on Travis-CI

[zbeekman]

It's nearly impossible to trigger this failure intentionally. It only happens intermittently, and I haven't witnessed it on a local machine recently. Here is the failure on Travis: https://travis-ci.org/zbeekman/opencoarrays/jobs/116974353#L433

[zbeekman]

@afanfa suggested overloading a single node with extra MPI ranks/CAF images than cores available, e.g. 16 for two cores... I can also do this inside of a vagrant virtual machine or a docker image that will somewhat match travis-ci environment...

[zbeekman]

@afanfa I have confirmed this bug on multiple systems by overloading them with more MPI ranks than cores. On my OS X machine, I see these with relatively high frequency if I run with 32 cores.