I requested this because overall latency does make a big difference. I have absolutely no idea what the latency cost of different Blocks is, why should I?
What I have read Cliff saying is that Pitch algorithms incur latency, and some of the reason some things don't get implemented on the FM9 vs the Axe III is related to latency impact which isn't quite as much of a problem on the Axe.
Regarding the 2 milliseconds jibe, if you think that 2 milliseconds doesn't make a difference, how about 5? The Headrush Prime was at between 8-10 milliseconds, and even higher for intensive Presets and a LOT of people were noticing it and complaining about it, so much so in fact, that Headrush supposedly released a firmware update to attempt to address some of the latency issues.
I haven't noticed latency particularly on the FM9, and reportedly it's somewhere around 5-6ms BUT if it does increase under certain circumstances to say 8 or 9ms, trust me you will notice when playing. I know this because the latency differences for various settings on my audio interface are noticeable when I'm playing guitar when the differences are as little as 5ms. If you search "human detectable latency in audio" you'll find this confirmed in enough studies to make it credible.
For me it would be useful because I hate it when I can detect it when I'm playing. I suppose you could argue that since it hasn't been a problem so far why do I care - it's just useful to know when I might be pushing the limits of what I'm happy with, similar to CPU. There are a number of synth plugins that I use which indicate current latency and I find it helpful when I'm working with modular stuff, where each component can have a significant impact on the overall latency - especially when they are 3rd party components.
If the FM9 could never really exceed 6-7ms then yeah it wouldn't be worth it. Do we know that?