As someone who's deployed several industrial RFID systems: I doubt we'll ever see an RFID tag on every item in Walmart.
The first reason is the cost of tags: Even the cheapest EPC Gen2 RFID tags cost more than 1 cent - and nobody selling a 20 cent can of beans is going to pay that. Barcodes, on the other hand, come free with the product's printed packaging.
And the tag's performance varies - the super-low-price tags might work fine for cardboard packages, but not work on metal cans, for example, as sticking a radio antenna to a bit of metal and having it work is a complex business.
The second reason is reliability: Even in a highly controlled environment, where every tag is on an identical item and you're willing to spend 20 cents per tag, RFID is hard to get right.
One thing that causes reliability problems is RFID is tough to test. Reflections and interference patterns mean you'll get areas where reading will fail, even when inches away reading works very well - and if you move metal in the environment around, it all changes and the nulls move.
And that's before you get into having a thousand different products from a hundred different suppliers, all of whom now have to learn this new RFID technology.
Another thing that leads to reliability problems is simple mechanical matters: It's tough for an RFID tag to survive being in between two metal cans that bump together.
And if you're hoping to do something like detecting thefts, a system that is merely 99.99% reliable is going to generate far more false alarms than it does real alarms.