If triple buffering is genuinely working then despite vsync the game can and will see frame rates higher than the refresh rate of the monitor. 1000 fps is perfectly possible if triple buffering is working. But if you are capped at 60 hz and vsync is on (no tears) then there is not true triple buffering. Not a single game on DirectX actually supports triple buffering, its simply a misnaming slightly marketing term, its being abused. Its using three buffers but the algorithm is wrong.
In DirectX you can setup a "flip queue" that has more than the usual 2 buffers (double buffering). However you can't set a strategy of flipping these buffers to allow one buffer to be constantly overwritten, they must be played through in order. Thus despite there being a third buffer the GPUs output will be limited to the frequency of the monitor. Much of the benefit of triple buffering comes from the ability to replace the oldest frame and to allow flipping in such a way that you can always get a newer frame. Without this strategy in place you just get a lot of latency and you don't eliminate the discrete 16/33 ms jumps present with the usual duouble buffering and vsync. In essence it is "triple buffered" but its not the real thing.
Games do not magically snap to 30 fps when they drop below 60 fps, because frames per second is an average. If all the frames are delivered in 16.6ms timing except for one frame that was delivered in 33.3ms then your tool will tell you that you got 59 fps, but you didn't get 59 evenly spaced frames, you got 58 perfect ones and 1 frame that took twice as long. We call this the 60/30 problem. Those little judders are very noticeable to those with experience.
-deferred rendering is essentially queuing of graphics commands for multiple frames by the CPU. Is this done by the application, D3D, or the graphics driver, or in the GPU?
Nope deferred rendering is a technique where some of the fixed pipeline functionality of the graphics card is skipped or nullified and then filled in using post processing. In essence the programmers of the game take on some amount of core functionality of the rendering of the scene and run it on the GPUs processing cores. Typically today this is done with lighting as the fixed function lighting on the GPU is not very good. It has nothing to do with the way in which the frames are delivered, it only has an impact on the GPU pipeline itself making some hardware features (AA) unavailable.
-flip queue/render ahead size. How many frames are 'deferred' as above maybe?
The buffers that the GPU renders into are organised into flip queues.
Render ahead is the size of the command queue, the commands before they go into the GPU.
-triple buffering. A general buffering technique, sure. But what does it apply to in D3DOverrider/RadeonPro?
It doesn't, these tools can't make directX do something it can't do. It is technically changing a game to use three buffers, but as I explain above without the algorithm it just adds a lot of latency and maybe a bit of smoothness.
-frame buffer. This is output buffer for the display and is essentially the last buffer in the graphics pipeline. This could be single (not usually), double, or triple buffered.
Triple buffered means more than just having 3 frame buffers however.