Any number of things - the two that immediately leap to mind being either a piece of corrosion in a heat-exchanger finally burning through to give a hole, or simply that the late night return flight was through cooler air so the cabin heather was turned up to a higher level (increating the amount of CO being pushed into the cabin). It's equally possible that the CO leak was present from the moment the engine started, but it too half an hour to build up to a fatal concentration.
The whole "it worked yesteday so it is strange that it broke today" argumnent that we hear so often is utterly inappropriate for complex systems - it assumes dice have a memory, which they don't (I checked).
PDR