Much as I enjoy various 4X and "strategy" games, the research systems in them are very much creatures of gameplay, not realism.
For example, the principles behind light-emitting diodes (LEDs) were discovered almost a century ago. And first commercial manufacturing was decades ago. Yet only recently they have become used for lighting.
Whereas in most games, "LED lighting" would be a "project" that starts in 2000 and ends in 2010 with a pile of components/units to be used. Fixed costs, start and end points and immediately usable applications
- forget realistic, they are not even trying for plausible.
One should not complain too much of course, a game that sacrifices too much gameplay would not be playable. Still, some attempts could be made toward suspension of disbelief, so important in entertainment.
But if research systems in games are to be more realistic, what is the real world "research system" like? There is a lot of information on various inventions and discoveries. Not so much on the mechanisms and patterns underlying them.
Here then are some general points on "real world research system", as I see them.
There are three parts.
1. Research itself, to discover patterns and rules that work in the world.
2. Application, to create ways of using those patterns and rules. Application is not always engineering as such, the advertising industry is all about choosing what to show by knowing how we see. And many other "soft" sciences have a precise and commercial application.
2.5. Applications can be primary, general building blocks that are built on Research.
Secondary, a combination of other Applications. For example smartphones
would consist of computing, flat panel screen, touch screen,
digital radio frequency transmit/receive and other Applications.
3. Designs. For example "main battle tanks" as an Application. And T-72 as specific design of this application.
While it is generally possible to show what Research an Application is derived from, the other way is generally not. Looking at some Research item, one cannot predict how the findings can be used before the research has happened. And often for a long time after as well.
Research is unpredictable. It tends to be punctuated with sudden discoveries between long periods of gradual gathering and refining of data. While the gradual parts are rarely newsworthy, they are necessary for discoveries to happen.
Everything in Research and Application is worked on at the same time, unless specifically forbidden. There is no "were done with refrigeration, now lets do satellites".
Applications, especially primary, often have an exponential curve of advance in effectiveness and/or efficiency. That is, the advancement is fixed relative to current state and accelerating in absolute values. The rate of advance tends to fall off as Application gets close to theoretical limits.
Applications that follow the pattern in previous point have a "period of maximum impact". A time when the gains in the Application result in large and significant changeover in Designs. And/or the appearance of many secondary Applications using it. For example chemical rocket engines had this period from the 50s to the 80s. The Designs before were not very practical. And the Designs made after have generally been towards cost reduction or other non-performance-related features.
A new Application can be invented even when there has been no change in the underlying Research. This is more likely for secondary Applications that often appear when primary Applications, that the new one uses, have reached some level of maturity.
Designs do have specific "start of development" and "end of use" points. Still the needed investment can be different than planned. Usually towards the upside.
There are many, many examples where Application is going forward
without problems. But a specific Design is no better or straight up worse than its
predecessors. The littoral combat ship (LCS) from USA naval forces being a recent example.