Terminating An Iteration

A few of us were discussing what would constitute a change in requirements that would cause a development iteration (or Sprint in Scrum terminology) to be abnormally terminated, ie what would be so serious (and/or costly) that you simply could not carry on with development ? Here's a recent example (kind of) from the aerospace industry.. Airbus's new A380 is facing production difficulties which does not bode well given the relative success (or progress) of its competitor, the Boeing 787 "Dreamliner" : the 787 is made of lighter composite materials that allow higher cabin pressure and humidity, adding to passengers’ comfort, as well as providing a wider layout for more seating. Airbus has therefore tried to counter the 787's relative success by "refactoring" its existing A330 aircraft to create a new A350 (in addition to continuing the A380) However, during development of the new A350, Airbus found out that the airlines actually want a "wider fuselage" - that would have been a nice requirement to know about at time of upgrade specification !!  This was always going to be a very expensive refactoring effort from A330 to A350 ($5bn), but not as costly as ignoring the change in requirements. Hey --what's an addition $5bn when all you have to lose is most of your market share to Boeing ? (OK, so their development cycles are a touch more than 30 days, and they are probably pretty spec/documentation heavy, amongst other non-agile activities, but at least they listened to their customers ! )