>I agree that, in general filtered grids have their problems but... I want to try to not have to rewrite a huge application. These grids derive fom classes used for filtering and incremental searches according to various indexes, throughout the system - in short this will be a heavy rewrite.
>
>It seems that the problem arises with the interaction between the grid, rushmore, and the data refreshing. When setting set refresh to 0,0 the problem disappeared. Also it seems to be happening more often on fast machines than in slower machines.
>Right now I disabled the refresh and setting only when I really need updated data. This seems to make things much better.
>Seeting the optimize property to .f. solves the problem, but also slows down the system, especially on slow networks,
>
>Maybe there is another way to get around this bug? A very small delay when refreshing the buffers, if this can be done?
>

Have you eliminated all the problems stemming from SMB2/Oplocks, which can hit you with Win7/Vista when working with dbf ?
Perhaps it is just another way showing the problem, which others not using filtered grids don't expierience as early as you do.
Could be some de-synching of cdx/data in the way SMB2 tries to buffer/cache.