Texture Matrix Implementations

Hi,

The game I'm currently working on uses the texture matrix often for things like detail maps and damage textures. I read in another topic that texture matrix implementations are problematic. I am not concerned with stack depth as we only ever push once but I am concerned with getting bad results. Everything seems to work fine on the nVidia parts we develop on ( TNT1s ), the ATI  in my G4 and a few other cards I've tried out. Is there a specific problem I should know about?

Thanks for any help,
Guy English