Also, the early ...a6 plan is flexible in that black can always go for the typical K-side fianchetto if he feels that white's aggressive intentions can be countered adequately.
I don't really agree with your assessment, insofar as it fits (or doesn't fit!) what people actually play.
I've started playing 2. ... a6 against 2. Nc3 but I've been playing it against 2. Nf3 for at least 25 years. I'm starting to get good at it and have developed theory quite a bit.
Normally, the move order would be 1. e4 ... c5 2. Nc3 ...a6 3. f4, if white is determined to play the GPA. If black continues with b5, white can pressurise it with a4 and play to disrupt black's development. It can become quite a melee. However, I prefer a more principled development as black and would therefore play 3. ... e6, intending ... d5. White can no longer play the positional line with Bb5. This Bb5 started to appear more often 20 years ago in club and tournament circuits in the UK and I decided early on that it's strong for white in the hands of a strongly positional player. So I avoid it.
So white's development against 2. ...a6 with 3. g3 is actually a complete change of plan, where white makes the correctly principled decision that g3 is the best way forward, against an a6 Sicilian. It's no longer a GPA, although white may try to play f4 before Nf3.
I occasionally castle long as black but by and large I prefer to castle kingside and open the queenside to use it as a pivot to attack white's kingside. Occasionally I castle kingside and advance the g and f pawns, if the K is safe enough on h8. Sometimes black's K is safe in the centre, especially if pawns are still on f7, e6 and d7 or d6.