First sighted here:
  https://mail.openjdk.java.net/pipermail/jmh-dev/2021-July/003325.html

This whole thing is guaranteed to fail if a long-running @Benchmark does not check interrupt status at all. Then the first thing that would check it and throw IE would be the control.preTearDown in warmdown loop.

Then it fails: the method runs for too long, so JMH starts to deliver interrupts, but method does not react and eventually exits on its own accord, at which point JMH infra catches IE and fails hard.