Yep, it seems you're the giant whose shoulders I'm standing upon...
(hope that I'm not too heavy !)
What I did add :
- The fact that all three are situational, so as to make none of them a no-brainer. Your initial proposal may be better on that point, if balanced right.
- The fact that techs are multiplicand, not additional bonus. That part seems extremely crucial to me, being at the core of the balance (one player is never so in advance in the race that nothing he does can be countered, and on the other hand being just on par with the enemy - or even a little behind in some situations - still allow some stealthy maneuvering if done particularly well).
- The Policies needed to activate the stealth. That part also is crucial, avoiding the "lost investment" effect of current stealth mechanism (if you invest in the right stealth path, you will always be able to use it against those of your enemies that did not choose to defend against it; if you invest in the wrong one, then it's your responsibility to have wasted all this Research).
Note that I did make the Stealth/Detection Policies for the same Path incompatible for fluff reasons, but I don't believe it's entirely necessary for balance; it's already difficult enough to get enough Military slots to defend (or attack) against all paths. Maybe Oberlus' proposition of having interactions between paths is better.
- The fact that both Detection and Stealth would rise in a slowly progressive manner. It actually stems from the fact that Policies are public, so my proposal was without any gameplay worth if switching Policies to match the enemy's choices was a viable way of dealing with his stealthy threat (since obviously such switching would be a no-brainer). But it has added interests : first it means priority on strategy over brute force (the one able to take a path early will have an advantage over the one who only reacts to it), and second it means than races are more subtle, the one ahead being both always in danger of being outpaced, and able to keep a temporary advance even when technologically obsolete (since any stealth has to be obsolete at some point at the end of the endgame).
My proposal works well with the different "non-binary" stealth systems, I think.
But it won't work anymore if these core points are removed.
Note also that it's not exactly a "rock-paper-scissor" mechanism, more a way to force strategic choices which have consequences that need to be thought out well and ahead of time, and to avoid that a temporary advantage spirals out of control (as whatever one does, he still would have weaknesses that he needs to defend against).