The SD1100 ranged around 60-100, while the SX230 ranged around 90-120 (once it caught a grey 80 just barely).
FWIW, the live view probably doesn't update at more than 30 fps (33ms between frames) and in my experience there's something like a 50ms delay between shoot_full being pressed and the shot actually being taken. So you are pretty close to the best case, although I guess it's possible that the SX230 is looking one frame behind the latest live view.
I've already been wondering whether I'd have better luck doing some sort of (long exposure?) intervelometer for lightning and I will have to experiment with that when the storms come.
This is definitely the way to go if you are shooting when it's dark, then you can just leave the shutter open for a long time and let the strikes pile up. An example
http://epod.usra.edu/blog/2012/02/lightning-display-on-ikaria-island-greece.html (they ran an intervalometer with 20sec exposures, and then stacked the results )
The nice thing about MD is that it will catch lighting in the day. Of course if you just ran an intervalometer in an active storm you'd get some too.
Mainly, I'm just surprised. I assumed this camera would be faster since its more powerful in general.
Quite a bit of the remaining MD don't really depend on raw processing power. The only thing that depends on processing power is actually looking at the live buffer. The sx230 has a higher resolution live view, which could negate the difference in processing speed. You can reduce the time cost of this by making the pixels step bigger setting bigger. For lighting, the whole scene is probably going to change, so you don't need a lot of precision. Changing this from 6 to 10 seemed on my d10 seemed to lower the minimum time from ~110 to ~90, although I didn't run a lot of tests.
The Y mode should be quickest, and is probably the best choice for lightning.
Masking some of the area would probably be OK too.
Hmmm, for lighting and similar, we could probably just make a mode that monitors Bv and doesn't even look at the live view.