1. start the most advanced rocket engines (full-flow staged combustion cycle engines [1])
2. ascend to 10km gradually shutting down (and throttling) engines to keep relatively constant speed while methane and oxygen are depleting
3. switch from vertical ascent to horizontal descent flight profile, and from 10km "fall" (or rather fall/glide) into the 30 meters target of the landing pad
4. reach soft landing (almost accomplished by the SN10)
BTW Roughly only the first two of those items are in the "normal" checklist of any other rocket company for an orbital rocket.
SpaceX is following somewhat Soviet rocket development approach with many testing, which is not a bad thing by itself. It is just different to what we are used with NASA/ULA/et al. SpaceX is also very open with their testing, you can literally see the rocket from 50 meters away, and its launch (from 8km) freely. We obviously have an observational bias here, while not knowing how many failures others have.
[1] https://en.wikipedia.org/wiki/Staged_combustion_cycle#Full-f...
Well, not. During SN9 flight, one engine could not be re-lit, so SN9 crash-landed. During SN10 flight, the two engines that were needed for landing were lit, but there was an imbalance in the fuel mixture, one engine burned noticeable more yellow than the other, and now this. You'd think they'd analyze telemetry data from flights and on-ground testing and improve their design, but visibly they don't.
And they are on notice already from the FAA for routinely blowing stuff up near a populated area.
The lessons from SN9 and SN10 are probably incorporated SN20 or even later version. (Or corresponding versions of Raptor engines.)
SN15 (the next launch) has a tweaked design of engines, but once again they are going to have multiple launches with these specific engines even if data from the first launch suggests some improvements. In fact, they had originally planned to do even more testing with this design (SN12,SN13,SN14 which had a bunch of work already done).
I wonder if while making Raptors some interesting techniques were created. For F-1 the result was so good the production engines routinely dampened instabilities, so Saturn-V didn't have this kind of problem in flight. Raptors seem to be less developed - of course, they have much higher chamber pressure and efficiency...
They're having some trouble restarting them on landing, because getting the fuel to flow properly with that dramatic flip in the air is a bit of a challenge, which nobody else has ever attempted.
I can't tell if you're trolling, or if you're serious? SpaceX does an incredible job of instrumenting their rockets. They do plenty of "on ground testing" of Raptor, but some things can only be tested during flight.
FAA gives launch approval knowing full well what can, and does, go wrong. It’s FAA’s job to deny on the slightest justification of risk notice.
Unlike the Soviet Union, SpaceX can afford to fund their version of the "N1" moon rocket: at around $1 million per Raptor engine (and 3 Raptors per Starship hop test) the R&D program is still relatively affordable, even with the failures.
This philosophy is in stark contrast to the SLS (Space Launch System) which after decades of development has only just done a static fire. Also SpaceX's philosophy around the Starship test program is focused on building a factory that can manufacture thousands of Starships vehicles a year.
But I agree on the spectrum between "analysis paralysis" and "reckless" SpaceX is arguably a bit too far towards reckless at the moment.
I hope apparently wasteful R&D program yields better results soon. They need to start testing rapid reuse for Starship to be able to meet its cost goals.