The demand last year was what, 600,000 in the first day? The 4S might be larger, since the folks who couldn't upgrade to 4 last time around are probably able to upgrade to the 4S, plus the new carrier, plus demand has been trending upwards anyway. But you have to ask, does it really make that much business sense to buy a whole slew of new servers and bandwidth you'll need once, and then never again (until the next iPad, at least)? Plus, it adds to the hype if they can brag about having so many customers it destroyed their servers.
Even if Apple and the cell carriers could plan for last year's traffic load, it's probably higher, and to figure out just how much higher in advance is rather difficult. Maybe they aim for 800,000 the first day, but will that happen right at midnight, or throughout the day? Is it really worth having enough to handle the midnight crowd if it tapers off quickly and gets much more reasonable an hour or two later? Plus, how many iPhones are even ready to ship? Stretching out the orders a bit may make it SEEM more fair, those who don't manage to get their orders in right away, therefore, won't feel so bummed that they don't get the very first batch.
It's easy to criticize, but these are all very difficult problems.