Hi, fluxer. Thanks for the thoughts. I would like to move to a multiple server model, but this does have some difficulties. One is performance (I've got to wait for N servers to respond instead of 1), and another is fragility (if I use 2 servers and they give different results... now what?) as well as ambiguity (If I query three servers and two never respond... now what?), as well as economics (each light server charges a fee for propagating transactions; if I have N lightservers, I have to pay N fees?). I haven't taken the time to come up with good answers to these questions, so the current version only supports one server. If we can come up with a good design, though, I'll be happy to implement a multiple server architecture.
As to the issue of delegates hosting servers, it's a great idea, but the light wallet has no way of knowing who the delegates are nor what they've published in their accounts. The light wallet would then need to trust a lightserver to give it accurate information about the delegates, which defeats the purpose.