If you look at the published video on YouTube of the explosion and go frame by frame, there are two events. The first is a bright flash that lasts a few frames and appears much larger than it actually is because it is both saturating the camera and illuminating the condensation clouds. You can see the illumination effect clearly in the first frame the flash appears as there are distinct shadows in the clouds. It's unclear to me whether this triggering event is electrical or chemical in nature, but I'm not an expert. Three observations can be made, however: (1) it is bright enough to cause lens flare in the camera which allows pinpointing its source despite the saturation (look for the X, carefully find its center -- you can do that very accurately -- and then back up a handful of frames; see that triangle thingy with a thin tail? That's what failed.) Then, (2) the initial flash is small and is followed almost immediately by a medium sized flash, and in turn that releases the fireball. Then, (3) the condensation clouds aren't moved by the explosion for about 12 frames until the fireball really starts to form, suggesting that the earlier flashes marked the release of lots of energy that may have been primarily radiation (light) rather than heat because they didn't expand the air enough for me to think of them as explosions. The video is 60 FPS, and the initial flash forms within one frame, so that's only 17 ms. The consdensation clouds don't start moving for 200 ms from the main explosion.
So we have one event that's exceedingly hot that triggers a second that's also exceedingly hot, that releases enough LOX to start the fireball. I'm thinking static discharge from the LOX filling.
One thing I don't understand, though, is that if you watch the fireball in slow motion, as the lower front heads toward the ground, there are seemingly waves passing through it. What are those? Additional shock fronts from tertiary explosions?