My opinion is that the PPC based implementation also haven't reached the limits of the processor yet.

I think, the first step could be something that more ABoxs are executed in the same time. It could be fine, because all core could be used to execute an ABox, on independent memory areas and a crashing app would only freeze it's executing ABox and not the whole system. I don't know is this feasible or not... But could be great!

The change to x86/x64/ARM will suffer from the same problems: Memory limits, freezings, etc.