You are confusing many different problems and coming to the wrong conclusion. Designing the shape of the F-117 required

high resolution 3D fluid dynamics simulations since the faceted shape required by the initial concept for stealth significantly deteriorated

and stability

The design of the F-117 was optimised for stealth which meant specific precise shape design... the computers of the 1970s couldn't cope with the enormous number of data points from a smooth curve surface, so the designs were made up of flat faceted polygons... like early computer games drew aircraft in flight simulators in polygons. the more polygons the more realistic shape but you would need millions of polygons to create smooth curves and they simply did not have such detailed models, so the computer model had flat faceted polygons and the final aircraft had the same.

By the time they were designing the B-2 and F-22 computer processing power had improved dramatically and curved shapes could be processed using what we called ray tracing... except instead of ray tracing light it ray traced radar waves to determine the radar view of the aircraft.

Obviously another aspect was precision in construction... I rather doubt they would have been able to produce aircraft with the curves to the accuracy required to actually make them stealthy when they made the F-117 either.

The cavitation noise of a submarine propeller is a problem amenable to asymptotic perturbation expansions

and revolves around the characteristics of the boundary layer flow. It is simply a different problem from the F-117 flight

characteristics.

Actually exactly the same... the difference being with the F-117 they were looking at drag and lift from the air flow so the damn thing would actually fly, while with propellers the key was ensuring a smooth water flow without separation and cavitation which in addition to making noise can actually cause serious damage to the propeller.

Also, the USSR had sufficiently powerful computer systems during the 1970s to actually do fluid dynamical simulations using

the Navier-Stokes equations. And the know-how on the applied math front meant that they could come up with approximations

to the governing equations in the boundary layer and use them in their model code. They did not need to run the equivalent of

climate GCMs at high resolution to model the flow around the propeller blades. They could even reduce the dimensionality of the

problem from 3D to 2D using the right coordinate system and slicing the hydrofoil.

All perfectly true, but if you don't have the machine tool with the level of precision to make that exact shape in 3 dimensions then it means nothing at all.

And how was some Toshiba NC machine supposed to solve any of the above problems? By definition it requires a part design

a priori. My statement was rather clear, once you know the shape you can build the part without NC machinery. Complex

parts were built long before NC tools. Toshiba "solved the USSR's sub noise problem" is a ludicrous fairy tale.

Making the calculations is not the same as being able to machine the product.

In the case of stealth the Americans knew full well what they wanted as a result but had to guess and model and test and then guess again and model and test which was an incredibly slow expensive way to do things. then they came across a paper by a Soviet citizen that developed the mathematics model needed to test using numbers and angles. this meant you could create computer models and test thousands of them per night and then produce prototypes of the most promising designs which can then be tested.

Having a machine that will create the propeller they could design for the Russians was like having the maths to rapidly prototype and test for stealth for the US.

Or if you like the Japanese selling a machine to the Soviets for sub propellers was like the Soviets selling the Titanium to the US to make the SR-71 and all the other products they produced in titanium.