On my i5 HD4000 I achieved 59-60 fps with 20 dropped frames and 11% cpu. I have to say that it's quite a benign test - which is why desktop browsers have achieved 60 fps for the last 4 years! I didn't notice any particle effects and the number of objects on screen never exceeded a couple of hundred; there's not much really much demand on the cpu which, on my machine, leads me to guess that the logic runs to only a couple of hundred events.
Most of the frame drops appeared associated with the creation of new instances of objects that aren't created on the layout start; this could be mitigated by creating an instance of everything at the start of the layout - none of them were likely caused by game logic. So IMO, it's a non-optimised and non-demanding performance test - probably good for the state of html5 a few years ago, but not really representative of a bullet-hell style of game that it's trying to emulate.
Don't get me wrong - c2 and html5 have come a long way since I bought r80 - and it's great to see Ashley's forecast of mobile platform html5 improvements coming to fruition. What would be more interesting for me would be to see if sound effects can be triggered in time to touch events across a broad spectrum of mobile hardware...