I think it's a bit more complex than that. This is the accuracy of a point detection. For a SAM system, you need to shoot the missile roughly where the plane is going - it's moving pretty fast. So your estimate of velocity will be vastly lower quality than the point estimate of position.

You also need to update the missile's course using updates on the plane's velocity changes - so again, you need even greater accuracy on the radar.

Finally, because it's all a lot more nebulous, it is easier for the plane to take evasive action. It's generally a lot easier to detect a radar / missile launch than it is to detect the airplane. With radar countermeasures in place, the SAM radar really has an uphill struggle to overcome.

I'm not saying it's not doable (though not too long ago I remember reading that these radars are just not accurate enough to guide a missile), merely that these dry numbers don't necessarily convey the full picture.

Not really. New AESA radars are able to home on countermeasures, so if the F-35 decides to use ECM or jamming, then all the better, the missile will be able to discriminate when it gets close.

The F-35 is going to be going at around 300m/s, and the time for the missile to get there from 60km away is going to be around 100s. You should be able to integrate position to get a +- 10% estimate of velocity with our 300m error, then. In reality a lot of the error isn't stochastic either so you can get a lower velocity error.

Again you don't need the radar to guide a missile. You just need to get the missile close enough for it to pick up the F-35.

If you read an article that said that the VHF radars can't guide the missiles, they are definitely right. But the Russians and Chinese don't plan on that. Their missiles now have on-board guidance, they just need to get close enough for their own guidance systems to pick up the target. That means that instead of needing 10m of accuracy, which I agree with those articles the VHF radars cannot do, you only need to get within a kilometer.

I don’t want to labour the point as I’m no expert, but is it enough to be close enough to the target and lock on? The rocket has some amount of kinetic energy for manoeuvering, and presumably less at the terminal stage. If the initial radar fix puts it in the wrong position/velocity relative to the target, can it make up for it with its own late lock-on?

I understood this is why you need that precise radar fix: you need a pretty good idea of the target trajectory so the chasing missile is not only in the right place, but also flying at an advantageous angle relative to its target.

As for on-missile guidance, I thought that was a thing since a long time ago? The radar guides initially, then the final stage is done by the rocket. Is there something newer/different now?

Until pretty recently, missile used semi-active guidance. Basically, they would have a receptor that would detect the radiation scattered back from radar. The issue in this case is that the signal would not be much more precise or even usable due to the low frequency. So active guidance fully from the missile is needed.

Ground based missiles are quite different from A2A missiles in that they are much larger and have much more kinetic energy to spare.

Maneuvering always kills kinetic energy though. So what these missiles do is that they gain altitude before maneuvering, reducing speed and drag, and once the maneuver is done (lock acquired), they maneuver and gain a lot of kinetic energy by coming back down.

There is basically zero chance a ground based missile fired from 50-60 km will lack energy even if it has to maneuver a few hundred meters to correct for inaccuracy.

Could you explain how a radar system is able to determine the direction/velocity of a track?

If you have multiple radars, you can use the doppler effect. You can also differentiate position numerically. Or you can do both!