The mentioned Java traceback is there even for OK cases.
Unfortunately, I don't know the exact reason right now. It may be related to the region of Azure hosting and for example, countermeasures for mining through GH Actions(performance downgrade). I run exactly the same CI on my account without failures: https://dev.azure.com/slev0538/slev_/_build/results?buildId=101&view=res... https://dev.azure.com/slev0538/slev_/_build/results?buildId=100&view=res...
Workaround: if test env's memory limit(sum of limits) doesn't exceed provided by Azure VM(7Gb+4Gb swap), then tests are OK:
https://github.com/freeipa/freeipa/pull/5656/commits/3f0308c275fc07a1cec87f3...
https://dev.azure.com/freeipa/freeipa/_build/results?buildId=8780&view=r...