UVR should queue processing requests because I often need to process the same sample with 10 different models before one of them turns up a good result and it's annoying to have to wait for each to finish before processing the next one. Also, the name of the model should be in the output filename so I can keep track of which one consistently gives good results.