My primary issue with software arrangers, especially those that use VSTi's, is the issue of having enough horsepower and HD I/O when running at latencies that are in the hardware range.

My G70 feels exceptionally crisp and immediate. It can stack five stereo voices up on top of each other, and play very fast repeated chords without a trace of flamming. I honestly feel like it is down in the 1-1.5ms total latency area (I use VSTi's a lot, so I'm pretty used to dickin' with various latency settings).

Now, I'm sorry, but I have yet to see a computer music system capable of staying reliable at those kinds of latencies, trying to generate 128 voices or more in realtime, while at the same time, generating automatic accompaniment in crisp response to live input. Especially when you start to use the better, more complex VSTi's available, nowadays.

The door on a hardware arranger is always pegged at the polyphony maximum, at exactly the same latency as the slightest use of it. But software latencies and capabilities are dynamic. The heftier the VSTi, the lower the voice count, or the higher the latency. And this changes all the time! Some songs won't tax it, some VSTi's won't tax it, some will. But you won't know you've hit the wall until it comes crashing down,

That is something a live performer cannot accept. At least, not this one...