I downloaded your files. The name for the .xmp file for the adjustment brush test image do not match, so the .xmp didn't get pulled in when I imported.
Remedied that, and performed the tests you requested.
The "moss" file you mentioned which appeared to not have any local adjustments would render a 1:1 preview in about 2.5 seconds. Probably the more useful test is moving from Library to Develop, as LR renders a new working preview when you do that. so the test is moving to develop then zooming 1:1 while in develop. The first step was pretty much instant, the second would depend a little on how quickly you did it ... if you waited a second or two then zooming to 1:1 was pretty much zoom then it becomes sharp.
The adjustment brush file does slow things down to a crawl. It takes about 16 to 18 seconds to generate a 1:1 preview, and moving into develop takes about 5 or 6 seconds to generate the screen size preview. Zooming 1:1 depends on how quickly you do it at this point, as LR is preparing the preview in the background. If you do it immediately it takes about 15 seconds, if you wait that long then it becomes instant.
But move to the brush tool and start trying to work, you soon feel like slitting your wrists. It appears whatever is going on doesn't thread very well. I'm also not seeing LR use the virtual cores so about the max I would get out of the CPU is 600% (where 1200% is the total possible with a 12 core machine). Doing these tasks however rarely use even 6 cores, and only pushes a couple of them past 50%.
So whether it could be better or not I do not know, and the question is does it really need to be better because this amount if intricate adjustment brush work isn't something LR really is built to do (at least so far). Don't take this personally but I'm not sure why anyone would do this amount of masking with the adjustment brush, since you could do it in a fraction of the time with far more control in Photoshop. I use the adjustment brush quite a bit, but nothing ever this complex, and I've never really had an issue where things slowed enough to make me feel the program was laggy.
I've heard Adobe is working to take advantage of the dual GPU's, perhaps they are also looking at ways to more effectively use multiple cores. Probably easier said than done ... the challenge is threading out the tasks efficiently.
Interestingly, Capture One 7 seems to be pretty efficient, using all 12 cores (6 real and 6 virtual) pretty effectively, getting 900% cpu usage on some things.
edit: After doing all of this, I checked the App Nap setting for LR, and found I hadn't disabled that. (Get Info on the program file in the finder). By preventing App nap, things improved quite a bit. Doing your test took only 11 to 12 seconds to generate the 1:1 when in Library, and I noticed that LR was using the cores much more effectively. Before it would use mainly the real cores, and each core being used would be progressively used less, so by core 6 it was probably only using about 10%. After disabling the app nap, it would use all 6 real cores almost the same and all the way to 70 or 80%, and the virtual cores would kick in as well to 30 or 40%. Before selecting the adjustment brush tool before took what seemed forever before you could actually select the adjustment brush you wanted to edit, and making changes was very sluggish. After disabling app nap while still a little sluggish, it was substantially better, and actually adding brush strokes to the mask was bearable.
If using Mavericks with any processor intensive application, disabling app nap is a good idea.
My specs:
Model Name: Mac Pro
Model Identifier: MacPro6,1
Processor Name: 6-Core Intel Xeon E5
Processor Speed: 3.5 GHz
Number of Processors: 1
Total Number of Cores: 6
L2 Cache (per Core): 256 KB
L3 Cache: 12 MB
Memory: 64 GB
