But that wasn't the main problem with creating a new jet, which has a myriad of other issues before certification.
That one's a killer. I was once a project manager at an aerospace company developing a new processor chipset for space applications. Our government customer would run out of money near the end of every fiscal year, forcing us to shut down. Then a few months later they would secure funds and ask us to restart. By then most of my staff would have been reassigned to other projects. It was stupidly inefficient and caused costs to balloon.
I don't know how much it applies, but sometimes evolving something requires switching out components for other ones, which are more expensive. Not everything scales. Marketing isn't the only reason why 50% more capability often costs 100% more. I imagine with rocketry, you're often operating at the edge of the capability for certain materials and approaches.
H-1: https://en.wikpedia.org/wiki/Rocketdyne_H-1
Merlin: https://en.wikipedia.org/wiki/SpaceX_Merlin
The Merlin has considerably higher specific impulse and thrust/weight ratios. It also has very fast throttling, a necessity for vertical landing. And, it's remarkably cheap (less than $1M per engine). I don't have the figure for the inflation-adjusted cost of the H-1, but I suspect it was much more expensive.
Is this level of optimisation really needed? I thought most of the landing was having enough fuel margin and then running a PID control algorithm during the landing burn.
https://www.nap.edu/read/23659/chapter/10
https://www.nap.edu/download/23659
"SpaceX uses CVXGEN (Mattingley and Boyd 2012) to generate customized flight code, which enables very high-speed onboard convex optimization."
This is actually a great example of a true NASA spinoff, btw. Lars Blackmore was at JPL before moving to SpaceX. Landing on Mars and landing a first stage back on Earth aren't the same problem, but they're close enough for the work to transfer.
The tanks on Falcon are welded using Friction Stir Welding, which was invented in the early 1990s.
The landing algorithm for the first stage uses convex optimization algorithms based on interior point methods, which were not available in the 1960s.
However, something very much like an expendable F9 could probably have been built in the 1980s. Simply evolving the Saturn 1B (which was cheaper, per lb of payload to orbit, than the Saturn V) could have easily beaten the Shuttle's economic performance.