Later I decided to see which viewer really was the slowest by semi-scientifically checking the system resources usage of each under the same circumstances. I cleared caches, rebooted the computer, and started a fresh session of each viewer in turn while watching the Activity Monitor that I have running. I reached a few subjective conclusions right away. First, Kirsten's seems to load images more slowly. Imprudence tends to leave textures gray longer, particularly alpha textures. Phoenix makes me wonder how safe it is to use every time it starts. The Official Viewer freezes upon start up one in three times.
Each viewer continually uses up RAM as time passes, even if actual activity is minimal (I stood in a poorly populated area while wearing very little, draw distance set to 130m, and a mix of prim types around). And each uses more CPU time as the session length increases but they all top out at about 60% maximum, occasionally dropping lower. Kirsten's used between 43% and 63% over all, the broadest range and the lowest minimum. Interestingly, the Phoenix viewer used only 9 processor threads while the Kirsten's viewer used 12 threads.
Practically, all the viewers are about the same. How slow one is depends on how long it has been open, what is within my draw distance at the time, what else is running on my computer or network, and how cranky I am (the latter being measured by chocolate and Mt. Dew levels).
All this testing helped me to decide that rather than test the Imprudence viewer for a week, I'm going to run Kirsten's for a week. I missed the v2.x interface it uses and it has some new features found only in the Official Viewer's beta (Web), like logging in to a Favorited location and the new Web-based profiles. I'll keep you updated.
By the way, the pic in this post as nothing to do with the topic. It just seems fitting.