I am a bit mixed on this debate. I truly think it could go both ways. Obviously the fact that Amazon, tried and stopped this practice says something huge. Amazon with their 12% conversion rate has been doing things right from the jump. But.......there could be certain circumstances where something like this could work. I know that for myself, we operate niche websites, and these websites offer durable goods, in which our customer will probably not come back to us for quite sometime. So where most people try to calculate the "lifetime value" of a customer, for us the "life time" is equal to a mosquito's life expectancy. We dont really expect our customers to come back to us, because the products life expentacy is so great. In addition, I could see other variable's work really well with this mod. such as a really low conversion rate. If your conversion rate is so damn low, you might be able to salvage a lot of customers in this fashion. So I just dont think anyone can say with any conviction that this mod will or wont work. You really need to think for yourself, and about your particular situation. Try to calculate "life time value", Harvard Business Review has a free calculator that can be found here:
http://hbswk.hbs.edu/archive/1436.html
Also, they dedicated a whole issue to this subject not too long ago.
You can also read a bit about the calculation on wikipedia:
http://en.wikipedia.org/wiki/Customer_lifetime_value
I would be interested in hearing any hard data anyone might have regarding this modification.
Also, some of you might be interested in the following research brief:
http://www.marketingexperiments.com/improving-website-conversion/abandoned-order-recovery.html
-DS