that doesn't say that it has to work with another vendor's gpu. afaik, you can still use 2 nvidia cards and run one as dedicated for physx. in fact, this is what i am planning to do when i play batman... i'm going to disable sli and force one card for physx. from what i've been reading, this improves performance a lot.
i don't see why there has to be a technical limitation preventing it. nvidia has no obligation to support the feature in a multi vendor gpu configuration. there is no technical limitation stopping you from clocking an i7 to 4.0ghz, but you will not get support from intel after doing so either.
the way i see it is, win7 has thrown a spanner into the works by allowing multiple vendor gpus to run. nvidia reportedly spent 150 million dollars to buy out aegia, and since then have agressively promoted physx... branding it, spending money to get it working on top games and what not. from a business perspective, it doesn't make sense for them to turn around and offer the advantage of using physx to their competitor with a 60$ second hand card.
let's not forget, supporting physx in a config like this would be an ongoing process. of course, i don't have any knowledge of this, but i would assume that running 2 different vendor cards on the same os would mean that the both gpu drivers would have to interact in some particular way. if something goes wrong with the physx, nvidia has to fix it or users would just get a negative impression of the company/brand. and otherwise, they would have to continue spending money and time supporting physx in this particular configuration, where they are hardly earning any money in the first place. this is just lose-lose for nvidia, and this is why they actively removed support for this feature, instead of just saying that it's not supported and leaving it like that.
and i really do believe that they would have maintained this support if they thought it would affect a large enough market. the one advantage that they get is that physx gains mindshare like this and possibly more devs would start using it. but they must have estimated that the people who this would affect was not worth the investment.
the users who really want physx have some options... either buy an nv gpu, or use the secondary nv gpu alone, or use the older drivers where it still works. it does screw over the people who specifically bought an nv gpu for physx, and that sucks. having an old gpu and expecting that to run doesn't count as being screwed over though. it's definitely not the best outcome for ati users but anyone who understands how business works will understand why nvidia is doing this.