I'm puzzled by the concentration on fuel problems as the most likely cause of this accident. I'm a software engineer and not an aeronautical one, but it seems to me that the failure of both the engines to respond to either autothrottle or manual throttle input is just as likely (if not more likely) to be caused by a fault in some common hardware/software subsystem. I started doing a bit of digging around to try to find some details of the architecture of the systems on the 777, and found some interesting references which I've given below. There is at least one large, centralised component that has a part to play in thrust management - the AIMS (Airplane Information Management System). I'm not saying that the AIMS *is* the cause of the problem, just that there *are* common subsystems involved in managing the engines.
BOEING 777 AIRPLANE INFORMATION MANAGEMENT SYSTEM OPERATIONAL EXPERIENCE
http://ieeexplore.ieee.org/iel3/5023...rnumber=635042
The 777 AIMS combines primary flight display, navigation display, EIC.AS display, flight planning, navigation, performance management, guidance, thrust management, central maintenance, airplane and engine monitoring, communications management, digital flight data acquisition, and data conversion gateway functionality into a single integrated system.
[...]
Thrust Management
The thrust management function sends commands to the autothrottle servo motors to move the airplane throttle levers, displays the thrust limit and autothrottle mode, and sends trim commands to the electronic engine controllers.
[...]
AIMS contains more than 600,000 lines of software
Developing the 777 Airplane Information Management System (AIMS): A View from Program Start to One Year of Service
http://ieeexplore.ieee.org/iel4/7/12850/00588382.pdf
The robust partitioning provided by the architecture allows applications to use common resources without any adverse interactions. This is achieved through a combination of memory management and deterministic scheduling of application software execution. Memory is allocated before run time, and only one application partition is given write-access to any given page of memory. Scheduling of processor resources for each application is also done before run time, and is controlled by a set of tables loaded onto each CPM and IOM in the cabinet.
I'm sure the system has been written extremely carefully and tested thoroughly, but there's always the potential for bugs in any piece of software.