If, under a specific research system, you find that replayability suffers because you keep picking the same tech path, that's your own damn fault for never mixing it up and you shouldn't bitch about how it's the game's fault for offering less replay value.
If your argument is "but but one specific path is clearly optimal", the following counterarguments apply:
1) In a game where there are a lot of viable ways to win or even ways to fight wars, how can there even BE an "always optimal path"? You should have to adjust your research picks based on what kind of map you are playing on and how exactly your opponents are playing. If you don't have to, then the game has design issues that go far beyond how tech research is determined and this whole debate's moot.
For instance, if it's due to there being a select few overpowered techs that are the best for any situation (lolsupplycrawlers), then guess what? Under blind research, those who roll those techs early crush those that don't. AWESOME. At least if you play SMAC with blind research off you can just choose to delay/skip those, or make an agreement in an MP game stating "no, this game isn't going to boil down to who gets SCs first. No using them.". With it on, someone's going to roll them first and dominate. (BTW I've never actually played SMAC MP so I don't know if SCs are really as overpowered as I'm making them out to be in MP, but they're a common source of BAWWWW from MP veterans so I'm running with it. And they do rape in SP, I can say that much)
If it's because of a lack of variety in options? Well, the replayability boat already sailed in that case. And then hit a glacier and burst into flames and sunk.
If you're assuming elemental will suffer from these kinds of problems, you have a pretty bleak outlook on how it'll turn out and should probably just skip it.
2) For any given situation, your "100% optimal tech path" is, 99% guaranteed, not. It might be the one you're most comfortable with at the time, but not the most powerful if played right. Experiment. Try different strategies in different games. You'll probably find something that works better for you.
3) If you care that damn much about choosing the most efficient tech path, why would you ever support an option that completely forbids you from playing in the most efficient way possible?
Realism for its own sake is never good design, it also needs to benefit the game in some way. The horn pro-blind-research people seem to be tooting the most is "it adds replayability", and I just don't see it at all.
This "you can choose to narrow your research but you take a penalty for it" idea sounds like the bad kind of compromise - one of those things that would half-assedly try to please all camps and end up leaving everyone only partially satisifed. This is worsened by the fact that a big part of the anti-specific-research camp, the part that thinks it hurts replayability, shouldn't even exist.
Stick with the traditional research model and let those people who complain about how it makes a game less replayable learn to discover the solution to that problem, because for any game that's well designed in other areas, it's there and it's damn obvious.
And for fuck's sake, people, stop insulting SMAC by referring to blind research as "SMAC-style". You make it sound like blind research was the only way to play that game. No, it was an optional ticker that defaulted to "off". And, one which was assumed to be off in most in-depth strategy discussions for the game.