Comment 3 for bug 1554572

I've been unable to reproduce this failure myself. And even in CI it doesn't happen very often. But it's not the only performance-related failure.

I would suggest moving such performance-sensitive acceptance tests into some mir_performance_tests2 and don't run it in CI, but then they cease to be helpful as a reliable CI check for regressions.

I think one possible way forward here is to just weaken the test so that it will pass more reliably even on slow machines. That means switching from a 59Hz generation rate to something a bit slower.