I don't know why the negative reaction. I've done the kind of analysis you've described many times and essentially been able to quickly identify such issues ove

I don't know why the negative reaction. I've done the kind of analysis you've described many times and essentially been able to quickly identify such issues over the years. We had a similar problem in Windows when we first implemented the Dynamic Fair Share thread scheduler. It took a couple months to have the right tooling to do a proper scheduler trace, but with that available the problem was better understood in a week. I eventually rewrote the scheduler component and added a control law to give better burstable behavior than the hard cap quota that this article seems to be describing.