An attempt at runtime synthesis using 32 AudioStreams firing off random explosion sounds was awful.  The problem is while sound latency is very very short, it's limited to frame start times, so no matter how randomized the explosion samples are, there's a hum at the frequency of the frame rate.  The other problem is there are no good internet samples of popcorn.  

This leaves longer, baked engine sounds played sequentially.  The japanese shuttle recording could be broken up into a number of loops for different stages of flight.  The problem is godot can't play sound effects sequentially.  They must be played in different AudioStreams as loops & faded in at just the right moment or played in a single AudioStream & interrupted at the right moment.  The interruption point is a little more forgiving in an explosion sound.

Physics simulation

In Godot, rigidbodies can only by moved by apply_impulse & apply_torque_impulse.  Trying to manually rotate_ a rigidbody makes it jitter as it fights the physics simulation.   Only Spatials can be moved manually, but this loses the automatic collision detection, bouncing, & integration.  The bouncing could enable interesting RUDs.

The center of mass is determined by the origin of the Rigidbody.  This is changed by having the model inside a spatial inside the rigidbody.  The spatial allows the rigidbody to slide around independent of the model.  You can actually change this during runtime to simulate fuel depletion.  

A true simulation of a 10 million pound rocket would be boring as batshit & real hard to control.  Kerbal space program already does this.  Arcade responsiveness with the apply_impulse functions takes a lot of fudging.   You have to integrate the forces on your own to reach a desired player input.

SpaceX has a way of coming & going.  There was a lot of hype after the falcon heavy launch, the 1st landing, the rapid pace of launches in 2017 & the BFR speeches.  Between the headlines, there was a lot of down time when no-one cared.  Now, they have a big slowdown in launch orders & the future has a serious oversupply of launches for very few large satellites.  Launching hundreds of cube sats hasn't panned out because of the challenge of synchronizing hundreds of customers.  The launching aspect of the starlink program is self funded.

Since this idea started, right after the falcon heavy launch back in April, it became clear that anything approaching even the poor quality of kerbal graphics is so complicated, it can never be finished by passive experiments a few times a month.  It has to be a full time commitment or it'll just be horribly behind the technology curve or outdone by someone else, which can make any experimenting a waste of time.

Of course, there have been Falcon 9 landing games & they all sucked terribly.    Guys throw up a model rocket with fake legs strapped on & crash it & it rounds the newsgroups.  Another BFR game has as much chance of being played by anyone else as someone else making an ultimate vlogging mic.