Walmart does A/B tests too, they just make the changes to a single store to hit that same 1% of its visitors
They'd have to make the changes to 10 stores to hit 1%. Just nitpicking, though :)
My understanding is that major walk-in retailers like Wal-Mart do multivariate testing to maximize the conclusions they can draw from simultaneous experiments. Apparently an approach called 'taguchi split testing' is especially valuable for evaluating combinations of changes, and is also applied in online situations that go beyond simple A/B testing.
Yes, if Wal-Mart had the magical power to allow 1% of its visitors to walk through the door to a completely different store, they would have the power to experiment with a lot of things that would arguably make the shopping experience better.
I have no doubt the Google A/B tests a lot of this stuff before rolling it out.