All fair points, Hyperion, and they tell me it's probably time to be more specific about some assumptions I realize I'm making.
1. Game vs. Sim: I'm assuming that Josh, when confronted by the inevitable choices between "is it fun?" and "is it similar to how our real world world (plausibly extended to science fiction)?" Josh will -- like virtually all game developers -- opt for "fun."
Where research is concerned, while there might be boom/bust cycles for particular techs, and/or for research generally (ideas that I quite like, personally), those feel more like simulationist features to me, where the fun is supposed to emerge indirectly from the interactions among cycles. But most LT players (probably excluding this bunch, but we're almost certainly not going to be representative) aren't even going to notice cycles; the only thing they'll care about is being able to maximize research results as quickly as possible -- i.e., pure mechanical gameplay.
I wouldn't be surprised if Josh & Co. seek a balance of some kind between these two flavors of fun. But I also won't be surprised if research winds up being implemented in a way that advantages mechanical fun: it delivers pleasurable hits of advancement on a
variable-ratio reward schedule through some calculation whether that calculation has any kind of real-world simulationist basis to it or not.
2. Thinking back on it, I think I recall Josh saying that research would be a late-game feature. So that, I think, affects the three different playing-time scenarios you described.
3. The alternative I think is being suggested to the three outcomes I described is "OK, just make research results multi-valued, and always balance benefits with penalties." That way technological power due to research stays at a sort of steady-state level. (Please correct me if I'm getting that wrong.)
If correct, is that something most LT players will find fun, where research on average does not deliver progress but only variation, where increased strength in one area is always compensated for by weakening the techs in some other field? Or will this be perceived as an obvious hack to avoid research discoveries becoming either too weak or too strong over time?
Ultimately, after reading carefully and thinking some, I'm not sure I've yet heard a direct counter to the list of outcomes I described being the only possible outcomes.
Having to play for hundreds of hours to begin to perceive variation in research achievement rates doesn't address the point of the actual gameplay value of a research discovery -- it's this value that I'm saying must either be capped, or must be decreased to avoid techs becoming overpowered, or are left to grow at a linear rate leading to some character's faction eventually becoming uber-powerful. In the first two cases, research becomes an irrelevant gameplay system; in the third, it becomes a game-ender. And the combination of all these outcomes means it's appropriate to ask: does it make sense to design and implement a mechanic that must have one of these outcomes? Will it be perceived as fun?
I lean toward "no," because I'm thinking that a reward feature that must either decline to irrelevancy or grow to be overpowering -- in other words, a gameplay mechanic whose play value will reach an endpoint --
cannot be a good fit for an open-ended game that can in theory be played forever. I'm thinking this is an inherent mismatch of features. And tweaking research to try to make it never end (so that it matches the rest of the gameplay) merely dooms it to static uninterestingness.
Again, I'm not saying any of this out of some preconceived and dogmatic dislike of research or tech trees. I like tech trees! I think it's a great part of the fun of 4X games.
BUT... 4X games are designed to end. So research can work in them.
LT, if it's still as described, is designed so that it doesn't have to end. That guarantees that for some non-zero number of players, they'll still be playing a universe long after research results have had to be capped or divided into meaninglessness, or that research results will eventually accumulate to a game-ending level of power for some character in the game.
That's why I ask: if a feature's going to end a game that's not supposed to end, or will become irrelevant, why implement it at all? Aren't there many other features with more long-lasting value that could be included instead?