I'm not a fan of the Image type for the vision stuff either. I keep getting caught out even though I know how it works.
A lot of vision stuff I do requires acquiring and image then creating various masks and applying them (often one after the other or in various combinations). The UI though, normally requires showing of original images and the results of the various stages of mask applications therefore you end up copying everywhere so as not to overwrite the originals or the intermediate results of a mask. It gets very messy
But my pet hate is that you cannot wire a VISA refnum to an event case like you can with DAQmx. But more generally, the "probe window" introduced with LV2009.