Isn't it enough that speculation is interesting?
This three words form a definite statement, and yet "interesting" is a matter of opinion, and based on his statement, "I don't see any point..." I would assume that in his opinion, it is not interesting.
Anyway, even though I also believe that speculation about limits is pretty pointless, I still sometimes find stuff like this interesting to think about.
2x2 Guimond:
I don't remember the exact numbers, but I know I did a rather noob-like movecount average with Guimond awhile back and it was under 16 moves. Certainly not an unreasonable number, and I'm sure when the cuber in question knows more about what he's doing than I did at the time, movecount average would actually be lower.
Step 1 is a small set of 16 "algs", an average of...what, 4 moves? Very easy to get fast at.
Step 2 is more often than not <R2, U> and only a few moves; again very easy to do very quickly.
Step 3 is PBL, a very small set of algs, and therefore very easy to practice a lot and get as fast as possible with it.
Judging by the above statements, potential turning speed in Guimond should certainly not present a problem. The only difficulty with Guimond (for some people - including myself because I haven't put real work into Guimond) is the matter of predicting enough of the solve during inspection to not have retarded pauses. I, for one, suck at Step 2 recognition, so if I can't predict the step 2 case during inspection, which I often can't because I'm retarded, the solve will suck. Luckily, this problem is pretty easily dealt with, even for one who is quite confident in his lack of ability to predict things in inspection. Since step 1 is an algorithmic step, with only 16 cases, somebody could work out a completely mechanical approach to predicting step 2. He could figure out and memorize a..."matrix" of sorts for each step 1 case to know where pieces would end up after the step 1 alg was executed. If the cuber in question got good at it, he could even use the exact same system to predict PBL as well (may not even be that difficult; I've never tried something like that myself), making Guimond as 1-look-able as EG. With the TPS some people can achieve, my un-addressable hypothesis would be that sub-2 averages with Guimond are just as possible as with EG.
