I have my A3 running at stock frequency with one APW3++ @ 115v (per my Kil-a-watt) with no problems since I got it. Why are you guys hooking up more PSUs to yours, if you are already using an APW3++? I don't even have the benefit of cold weather, with the high being in the 80s lately.
APW3++ are 1200W at 110V. Miner is about 1350W What you're doing may be working, but you're pushing the PSU outside the specs, and that is a risk.
Per Bitmain, the A3 is 1,186 watts DC. The power coming out of your wall is AC. The PSU converts AC to usable DC power, with some loss in the conversion process, leading to a higher draw in AC power at the wall than provided to the miner in DC. The only things one APW3++ can't solely power on 110v within specs at stock frequency, with its 1,200 watts DC max, are most S9s (11.5 TH/s batch excluded) and the T9. The APW3++ is operating within its specs, even if it is close to the top with the A3.
You are correct the page says that now. I'm pretty sure it said something different when I first checked, or maybe I got caught in people repeating that 1300W+ figure over and over. I know someone measured 1500W at the wall too (without overclock IIRC). In any case, if you're pulling less than about 1300W at the wall you should be OK.
I have 220V installation and most miners do because server PSUs usually give you more watts at 220V and you can take advantage of that. PSUs are usually more efficient at 220V too. Also, dealing with half the amps help with breakers, cables, etc...
The page does say that now, but as you rightly point out, draw at the wall is being reported much higher. Then again, maybe the March batch runs more efficiently? Makes me think of the D3, which is currently listed as being stock @ 15gh/s at the same 1200 wattage as the batch I got (which was listed and defaults @ 19gh/s). Not sure if the hardware is different, but considering that with BlissZ's firmware I can get 15gh/s out of mine @750...I trust my kill-o-watt way more than I trust Bitmain's numbers, especially when the listing isn't the same batch as the miner I'm using.
Half the amps is somewhat of a misnomer, you can run twice the amperage of a 220v intake breaker at 110v. The amperage on a breaker is a safety cut off, you still calculate the total wattage. Not really relevant either way, though, since 220s take up twice the space as 110s in the breaker box anyway...
And second vote in most cases for 220 (even though all my rigs are 110), PSUs and especially server PSUs run around 5% more efficiently (iirc) at 220. But at gold rating and above that discrepancy diminishes (or so I'm told, I can't currently do a comparison).