(2) rather than a 'default' 2000 Max Dimension, would not the maxium dimension of the Input image be a better choice? with the understanding for visual element "detail" as provided from the F1 Help...
(3) an Apply/OK image version does not have an "Untitled N" format in the frame name prior to saving it to indicate that it is an "unsaved" work space image.Step 1 - Resize
Before any other operations are performed, the input image is resized. This is done by specifying the maximum dimension (width for landscape or height for portrait) in pixels of the result image. Generally the image size is reduced in order to reduce the amount of detail in the image, but if the input image is already small, resizing it larger may also produce good results. The larger the output image, the more detail it will capture.
Slightly Off Topic... ...with a little bit more experimentation, I can see where this transform could potentially have un-intended "skin smoothing" and/or luma-chroma noise reduction possibilities.
...den...
Vista 32 Home Premium, SP2, latest version/updates
AMD Athlon 64 x2 6000+ processor
3 GB DDR2 dual channel RAM
2 MB ReadyBoost Cache
500GB hard drive with 100GB free space