There's some tricks you could do in DX10/11 for this, but I don't know of any in DX9.
The only hack I can think of is:
If you can guarantee that you've got no overdraw (two triangles covering the same pixel), which you can if you're doing a z-pre-pass, then you can render with additive blending and output black when you don't want to write data.
Yes, I had the same idea, but I'm not doing a z-pre-pass right now and I don't know if it would be too expensive. In any case, I'm pretty sure this solution would work, but I'll leave it as the last option
If you're using a pixel shader, you can create another technique that doesn't use the clip funtion and use it for those RTs.
Rememeber that clip doesn't return anything for that pixel, nor does it write any information to the back buffer or depth buffer. On newer graphic cards it stops the execution entirely.
That would work. But is it possible to use different techniques, for the different RT when using MRT?? If it is possible, could you give an example or a resource explaining how to do that? I've not been able to find any :S