Comment by zbentley

Interesting that it’s temporal (according to the article, you have around 10 microseconds before the signal-based preempter kicks in). How bad is performance if the load on the host is so high that double-preempting is common, I wonder? Or am I missing something and that question is not meaningful?