Steve's idea of creating the texture in default memory is good. I think you don't have to keep it there, either. You can create it, if creation fails scale it and create again, until you succeed. Then delete it and recreate as managed. As long as you don't create any default pool resources after that, I think things should work.

Another idea you can try is using a dynamic texture. There's a chance that'd be streamed directly from system RAM, instead of fully loaded onto the card, and therefore work (although more slowly than an on-card texture). That could allow you to use a larger texture than would fit on the card. It'd still be in the default pool, and therefore suffer from not being managed by Direct3D.

Last idea is to split that texture. Since I don't know how you use it, I can't tell if that's practical, but if you're only using a subset of the "frames" each frame and your shader will be able to access all you need even if they're not all in one texture, then splitting into several textures with different frames (with overlap, if necessary to get it to work right) will mean that you'll have smaller textures than Direct3D's texture management could work with, loading the ones you need.

BTW, as mentioned in another thread on the subject, reading "Resource Management Best Practices" in the DX docs is a good idea. Won't help greatly with your situation, but if you haven't read it, it's good material.