Initially, all our workunits are sent out to availible volunteers. Unfortunately, because of the low amount of workunits in this model, you were one who didn't recieve any. So, while you're waiting for workunits to download (which wont come because they have all been sent out) others will be crunching and returning results, in effect changing the percent complete.
Obviously, this is not optimal. We're currently working out a way that will allow workunits to repeatedly be sent out until all results are returned. The downside of this? Redundant data and heavy-traffic. Still, it's an avenue we're actively pursuing. Hope this explanation helps :)
As for the server connection error-- the server status page is cashed. There may have been a few moments of disconnect from our project servers, but since the error was resolved within the page's cashed period of time, the error would not appear. Now, if you're still experiencing that error, continually or regularly, there may be a seperate issue at hand.