>Peter Zijlstra started out the "state of the scheduler" session by noting that, on occasion, Con Kolivas surfaces with a new scheduler and people start sending in scheduler bugs. He would really like to short out part of the process and get problem reports regardless of Con's release schedule
Maybe the 'one CPU scheduler' in main branch policy is backfiring in a way?
It's much easy to see that there are scheduler bugs when you compare two schedulers than when you only have one..