As predicted in the PR, you're now stuck with a test that's broken (although the fix and the original unit test are still just fine). So you either have to not include in 5.0 a fix for a known bug (which appears to be what's happening), or spend time cleaning up my integration test that is broken in the new release (broken because there's no documentation for the test framework and I was stuck making up a solution that worked but was fragile for reasons I couldn't have known).