In general, for most components the cost in minerals scales proportionally with size.
I'm not about the cost in minerals, I'm about the cost in build points.
I can understand the concept of cross section perfectly fine, and at the same time believe that the fire controls and weapons in Aurora can have sufficient precision that cross section effects are much less important, even negligible, compared to extreme range, speed, evasion, and so forth which characterizes the Aurora universe.
Yet the basics of cross section concept is that range and speed are completely incapable to negate target's cross section effect. The only thing that IS capable to negate target's cross section - is impact's cross section. But in Aurora our damage model is inconsistent with a supposition, that impact's cross section is much bigger then target's cross section. While impact's cross section is less then (or comparable to) target's cross section - range and speed are only multipliers, relative to target's cross section in ANY equation of hit chance (well, higher math is about stochastic distributions, not sections, but it's only difference is that distributions are not plain, all other factors remaining nearly the same).
You can move your target within millions of km2 section of probable positions - and yet a target with 1m2 cross section will be still 100 times less likely to be hit comparing to a target with 100m2 cross section with the same cross section of probable positions (that is speed and time), if you are trying to hit it with an impact, that is less then 1m2. Your hit chance may become astronomically low, yet a target with 1m2 cross section will remain still 100 times less likely to be hit comparing to a target with 100m2 cross section. Your hit chance may become rounding to zero with your processing system, but it's irrelevant to those situations, where you have observable (and so not rounding to zero) hit chances.
Anyone who works with beam optics of any sort will tell you that this is utterly impossible
That's why my personal lore is that in Aurora combat space is a projective 2-dimensional space with compressed effective distances (i.e. Aether).
An alternative solution is to rename all relevant tech and branch names throughout the DB, but it's tough, while projective space lore is easy and amusing.
Though no math that I can imagine will negate cross section effect. It's much more fundamentally rooted concept - it cannot be outflanked by some imagination I can do, it have to be just "deunderstanded", and I just cannot do such a thing.
A quick Google search reveals the following:
For a F-35A the required amount of labor is around 40,000 man-hours, as of 2017/2018. Of course this has likely decreased since then but it is a good rough number.
For a Nimitz-class CVN, 40 million man-hours are required, a factor of 1,000x greater.
That's very different workhours - most of SY works are drastically more robust, needs more unskilled workforce (you can see it with your numbers - 1000 times more WF and still only 100 times more in wealth, and it's with scale effect). So what we have to balance - is some abstracted BPs.
It will open also a way to model scale effect with cycled operations (serial mass production).
It seems to be that mechanically, the "issue" is that fighter factories are much more efficient than shipyards because they can operate as an aggregate unit. Something similar to how GFTFs work would probably be more reasonable if the goal is to model real-life production more closely.
That's really very close problem: the same as with fighters vs warships we have a swing between nearly momentary start of production for small formations (so you have to remember to manually delay your production or stockpiles every time, if you want to RP, and if you are not - it will be a mess of forgotten tasks and stopped production), while bigger ones (like STO guns) are disproportionally slow and costly. With linear law for BP it will be nearly impossible to fix this - we'll have to jump from one shoulder of this swing to the opposite inevitably. With slower BP general law (smth like square root) it will be possible to easily fix the second end of this problem in the next iteration - by adding preliminary operations (~building prototypes and production lines as an abstracted, automated, yet tangible background). Thereafter we'll have a combination of not-so-slow build time for small-number large-scale objects and in the same time a slow start for serial mass production. It will be more consistent (one cost law for any production type) and much more player-friendly (united pools of industrial facilities without unnerving too-quick-ending small tasks).