There certainly are benefits of going to the gamesdb branch, as is would allow for much more flexibility in how ES presents its lists of games. Filtering, sorting, across platforms, per publisher,etc. These things are now completely out of the question.
What it does *not* necessarily mean though, is performance improvements over the current xml approach. It will need a strongly optimized implementation to even break even. Like Herb said much testing and discussions on how we would like to see this implemented.
I would argue that we could start the discussion already, and see if there is some kind of consensus on what improvements we need/expect for this significant step.