It doesn't take too long to load sample metadata, but it is a bit annoying, especially as the number of sample sets grows. The time is all network latency, because lots of relatively small CSV files are being retrieved. It would be possible to load these asynchronously, which would probably speed things up by an order of magnitude.