In this monster post, "Scott" discusses my Market-based Credit Card (MCC) idea along with Bill Me Later, a single institution version of MCC. I wasn't aware of Bill Me Later and there wasn't much information at their website, but Scott's description of their system sounds intriguing. With a single institution, many control and scalability problems dissolve to manageable technical issues. It will be interesting to see how their system do in the market.
One interesting aspect of MCC is that the system should work for any transaction amount. It can handle car and home loans as easily as buying a book at Amazon. Institutions bidding for car and home loans will be different from those bidding for $30 credit card purchase, but the cardholder doesn't see a different. The difference at the backend is that, for small purchases, institutions will consider only cardholder-specific risks and only in veto mode (approve by default unless the account is flagged). For large purchases, institutions will consider transaction-specific risks as well as cardholder-specific risks.
Another interesting aspect of MCC I failed to mention: there could be more than one market. You can have country or region specific markets to mirror political boundaries and industry specific markets (i.e. mortgages) to mirror industrial boundaries.