Are you talking about the exhaust plumes or the nozzles? I can't be bothered to watch that again.
My thinking goes like this: we're seeing a jet of non-reacted oxidizer at a fairly steady rate of flow just prior to the roll-over. We aren't seeing any odd flames following the bad engine, so there can't be any fuel reaching the combustion chamber or otherwise exiting the rocket body. That would mean that the fuel was commanded off or the flow is being stopped at the tank, the pump, or somewhere pre-chamber in an uncontrolled way. Presumably the flight computer won't allow either fuel or oxidizer into the chamber unless both lines are at pressure.
Because we see oxidizer, was it being released by command or by failure? How long would the system wait before cutting the feeds once it notices the failure to light? Does it have a "no matter what, run for 5 seconds and then start checking things" sequence?
That was an awesome spin, BTW.