SpaceX recovery

Yes, pretty good.

I wonder how reusable it really is, though? Also, the weight of the fuel used to land could otherwise have been used for the payload/second stage.

It was quite impressive how it managed to almost hover and manoeuvre itself into the “chopsticks”. By not taking its landing gear with it like the other smaller boosters they operate that saves weight and means more fuel available for deceleration.

Hovering is actually a hard problem.

The ship is almost out of fuel, so it's very light.

That is why some landings use the slam design. It's because the engines don't throttle well enough for perfectly proportional landing control.

The 33 engines aren't exactly the same. Only certain of the engines are candidates for landing. and there is enough redundancy, to handle failure conditions.

"Next booster will have 33 Raptor 2 engines, with 13 steering. Ship is being upgraded to 9 engines (3 sea-level gimbaling, 6 vacuum fixed) with increased propellant load."

A vacuum engine has a bigger bell on the bottom. You can see in the example, it wouldn't be easy to be swinging the one on the right around. A lotta mass. Lots of power to run the gimbal. The one on the left would be a lot easier to swing around.

The engines don't look like that any more, as there is a new version with reduced plumbing on the outside.

These are three sea-level engines, on pallets.

And most importantly, its not trying to climb out of Earths gravity well with a pack on its shoulders...

Directional control can achieve very smooth vectors sum thrust indeed. Viz a pair of rockets pointing sideways don't arrest the drop whereas a pair pointing straight down do