Proving to the network that you actually have a certain subset of blocks is an interesting problem, but I'm not sure that this scheme is that much of an improvement over being blind. If providing proofs for multiple IP addresses at once is seen as too taxing to be possible then it would make for a trivial denial of service vector, if it's not particularly difficult then it would be possible to expose a node with as many interfaces as you felt like same as before. To get anywhere near a decent picture of the network you're going to be asking a lot of peers for proofs at significant processing cost, and I'm not even sure if it's in their interests to supply them to you.

The non-listening portions of the network also make this a little unattractive. The bulk of all nodes don't expose listening ports at all, either by choice, lack of uPNP or by being behind a restrictive NAT. If you don't include them you lose a lot of the possible resources, if you do include them it's a mess of dynamic IP addresses and proofs that can only be obtained by the people they connect to.

I'm not convinced there is a solution to sybil nodes, isn't that what Bitcoin is an admission of in the first place?