It is possible to re-broadcast to a different server; it looks something like this:

webcam(s) --USB/VLC--> local PC --VLC--> fast network server <--VLC--> remote viewers using VLC

The local PC sends high-res data. The fast server has different options for limiting the bandwidth, which could include transcoding for lower frame rate, smaller image size, etc. I think my example isn’t really correct for limiting the bandwidth from the video codec: it is supposed to be around 16 kB/sec and it seems higher. It is also possible to have the server embed the video as Flash or WMV in a web page, it only takes a bit of web programming.

As for broadcasting to some video broadcasting service provider, in real-time, I’m not sure. Although frame grabs could be put up on any web host (flickr etc), and video captures could be uploaded to any video service if encoded properly (blip.tv etc). For my bio project, I am most concerned with frame rate and delay, because the robot arms are moving around and I need to see them in actual time. In other bio projects, frame rate probably doesn’t matter, and resolution is more important, so frame grabs might be the best bet there.

My solution is really meant for viewing by 1 remote person, possibly 2 (those running the experiment).

On the microscope+webcam front: I found proof-of-concept good results by literally scotch-taping a Sony PS3 Eye onto the neck of a $50 eBay 40-400X microscope, but I agree that some intermediary optics would be ideal.