I write software for use by high risk insurance carrier. There is coverage for just about anything. One of the inland marine insurance types might cover this. (possibly Electronic Data Protection or Valuable Papers)
Insurance won't insure for what they don't understand and build a risk model for. I can assure they won't understand something like this for a very long time.
I'm not sure they need to understand bitcoin specifically to build a risk model, wouldn't they just need general data on losses suffered by various internet hosts due to hacking attacks? Doesn't really matter exactly what is stolen as long as they have a corpus of data on the value of everything that is stolen in this manner. Bitcoin wallets probably fit somewhere in the payoff curve for that.
I've never heard of one, but there may very well be a market for that sort of thing. The problem with Bitcoins will be the legal jurisdictions involved, since these services are regulated nearly everywhere, right?