Seems to be failing over 2% of the time, which is high enough to be bothersome.

https://ci.wildfly.org/project.html?projectId=WF_PullRequest&buildTypeId=&tab=testDetails&testNameId=-7993376176519000819&order=TEST_STATUS_DESC&branch_WF_PullRequest=__all_branches__&itemsCount=50

My impression is this is testing pretty async stuff (metrics generated by requests propogating to a collector in a different process). So probably it should keep trying in a loop until a timeout has passed.