I'll dig through old papers. No promises though. If I find the slides I'll mail you the presenters' names.
Hey Gail, I was just wondering if you ever had a chance to find this info - I'm chasing down a performance issue with stats gathering on a weekend maintenance job and I'm highly suspicious that NUMA configuration is playing a big part in the equation.
On lower lane servers which have 2 - 4 NUMA nodes (distinct memory clerk IDs recognized by SQL Server), stats gathering runs normally using a backup from production - so I know its not "distribution" related. In my production lane it runs slowly but there's only one distinct memory node to work with.