successful != good
Amazon No Longer the Role Model for E-Commerce Design (Jakob Nielsen's Alertbox). Oddly enough, I never though Amazon's design was good. I think Amazon's success comes despite the terrible design - what they offer is just too good to pass up, so people tolerate the ten millions links per page and ever-shifting interface.
The same applies for eBay. Their site interface is awful, full of weird icons and odd navigation systems; not to mention their habit of sending 100% graphics email. But, the lure of snagging a bargain keeps people hooked long enough to learn how everything works. If that lure was any weaker, they'd give up. If you had to learn eBay's system just to pay your library fines, you'd probably give up and just drop in with some cash.
Basically, sales success does not equal good design. While a good design will definitely help sales, having good sales does not necessarily mean you have a good design. It could just mean you have a product people want or need enough put up with your design. I thought that was obvious, but since people continue aping Amazon, eBay etc, I guess it's not obvious.
Think of it another way. I've never encountered a VCR which had an easy, reliable timer record system; aka the "record the football while I'm at Grandma's birthday party" feature. There is no consistency - all VCRs are different, they label things in bizarre ways and for extra points some even make you swap between buttons on the machine and the remote during a simple process. Why do we put up with that? Because we really want to watch the football when we get home.
Motivated users will conquer bad usability, because they have to in order to fulfil their wants and needs. Give them a better option and they'll be off like a shot.