Even a $1,000 Amazon gift card was instantly credited to my wife's account.  She used it at Amazon before her payment to BitPay received even its first confirmation.  

But this just shows that zero-confirmations are actually very secure^* provided that the transaction has been accepted by many nodes in the network and that no double spends have been detected.  Double spending is very difficult and one can only succeed with a probability proportionate to how much global hash power one controls [https://bitcointalk.org/index.php?topic=465474.msg5151306#msg5151306].  

As I calculated above [https://bitcointalk.org/index.php?topic=465474.msg5151972#msg5151972], to succeed at double spending 6 - 12% of the time, one would need to purchase $24,000,000 of mining equipment [and risk having this forfieted by law enforcement if you got caught].  This is not a concern.  


^*The malleabily problem has revealed an edge-case that makes a small subset of zero-confirm transactions less reliable, namely unconfirmed transactions that contain as inputs the change outputs of other unconfirmed transactions while the network is under malleability attack.  This is the "malleability bug" that people are worried about--it is a real concern.  While the core developers work to eliminate malleability entirely, the work-around now being rolled out is to prohibit wallets from creating transactions from unconfirmed change.