From what I can tell, and I think you can see a bit in the video, when the grid fins ran out of hydraulic pressure they locked full in one direction.

The last burn this thing does is often called a 'suicide burn' as there are a couple complications to landing it. First, it only has enough fuel left for a short burn. Second, it's not possible to run the single engine they use to land it at a low enough thrust level that it can hover. i.e. 1 of 9 engines at the lowest possible thrust it can operate at is more than enough to accelerate the empty stage body upward. As a result they have to 'blip' it just before landing and cut at the exact time needed to reach 0-ish velocity at landing. This means the thing is moving _fast_ just before the video starts.

I'm guessing that with the grid fins locked in an extreme direction 24 seconds before getting to the barge they still had a lot of influence over the flight path since it was still moving fast. This pulled the top to an angle. The engine fired and the vectoring capability of the engine tried to toss it back over the ship but it just couldn't correct the lean before hitting the barge.

I really appreciate his matter of factness in these sorts of tweets.

Was probably the engine gimbaling and thrusting up to attempt to correct but was just unable to due to the fins not being around the whole descent.

Not only were the fins unable to move, the position they were stuck in was a very unhelpful position.

Yes, and could speculate that the main engine fired 'wrongly' because the control code didn't take into account that the fins were locked.

Hm, if I was the guy designing the control loop on that thing (not gonna happen in real world, but bear with me), I would definitely take into account the fins position - the real position, not the one I want them to be in.

Not that it would have helped anyway in this case.

I suspect that the code takes the fin's starting position into account in each planning cycle but also assumes that the fin will move as directed in the future.

No rotary encoders on those things?

To me, the video has an optical illusion effect, caused by the bright light from the engine. The rocket is not directly over the platform, it’s actually off the far corner by about 50-100 feet, it comes horizontally into the platform and crashes over it. Meaning it’s not close to the center of the platform and trying to accelerate off of it, it was laterally moving over it from the side and crashed the engine into the barge, which kicked it up over the platform.

Maybe they could improve the failure behaviour with some failsafe mechanism? *If the rocket is not within a certain envelope, gain as much distance to the pod as possible? Later when they want to have landing on land, a designated area for aborting could be used.