The project owner is the customer for the project. He or she approves the budget, and is also involved in iteration kick-offs and retrospectives (evaluation for an iteration). If at the end of an iteration the business case for the project dramatically changes, the project owner might consider stopping the project. In real life, this rarely ever happens.
I've experienced such a situation only once. We ran an agile project at a large reseller organization. During Propose and Scope we identified about 85 to 90 smart use cases, so the project had a fair size. The project owner decide to go forth with the project and we headed of enthusiastically. I just decorated the project room, order post-it for our dashboard and chairs with wheels to move around in.
It wasn't until the first iteration that we discovered that we missed one single line in the original customer documentation. This line stated that reports should be created for most of the main domain objects. We quickly added the new smart use cases to the model. When we were done - at 21:00 - we ended up with 145 smart use cases. Here the scope dramatically changed, so the project owner killed the project, as expected, during the retrospective at the end of the first iteration.
Sander Hoogendoorn