Which is to say at t-10 seconds when its pretty simple and routine to scrub, they do a massive comparison of sensor data and system status versus expected data and expected status. If that fails for any reason they scrub. It takes them a while to go through all of the data to figure out what data caused the scrub and then a while longer to figure out what caused the data to be "wrong".
My question though is what was that data anomaly? Also was it the same thing both times? Just curious really.