Big speed boost on larger grids by not always using the cell prototype duplicate machinery if no cell prototype was given. Cell prototypes are optionally assigned to row instances so that new cells are created using clones of the prototype (should you want), but most of the time this is not used.
It is still necessary to introduce a paging scheme so as to eek out more performance for very large result sets. Right now, it takes approximately 10 seconds to display a result set with 7,500 cells. A large grid, yes, but not unreasonably large. So much more work is needed.