I think if legacy paths are staying in the general way they currently exist, the sweet spot for me are ones that basically trick new players into playing well by giving them an extrinsic incentive to do what experienced players do intrinsically, and, by the same token, therefore don't require experienced players to go out their way to hit those bonuses.
What I mean by that is that completion of a given legacy path should not only come as a natural result of playing well for that particular type of game, but that doing so should set you up for future ages (beyond just giving you the legacy points). I think the best example of this in the game currently is the exploration science path. For a science game where you're ultimately really looking to max out your science and production, identifying and utilising high-adjacency tiles is a solid gameplan regardless. Five 40-yield tiles means you've made good use of your tiles, grown your cities, progressed through the science tree to get specialist limit increases, and all of those things contribute a lot to your empire's general strength going forward, either into a science win or to be adapted for something else. For an experienced player, they're probably doing that anyway, and therefore don't feel too railroaded into going out their way to get the legacy points. For a new player, meanwhile, the legacy points have the effect of jangling the keys in front of them to give them a more immediate, "gamier" reward to work towards, while at the same time guiding them towards taking action that will strengthen their empire in the longer run. It strikes a really nice balance between making the game more approachable for new players while remaining inobtrusive for those who know what they're doing. (I think antiquity culture and economic paths are pretty solid at this too; exploration economic isn't terrible either, at least in the sense of pushing you to expand effectively. And kind of just discount military from this whole discussion since it's kind of its own thing).
Contrast that with codices or relics, where it's just this weird arbitrary points system that only really serves to try and move you through the tree in question (though relics don't even do that very much). Yes, getting lots of codices requires a high science yield, but at that point why not just tie the legacy points to science per turn thresholds? There's nothing about codices that gives you more general strength going forwards (the "general strength" vs. "specific yield strength" distinction is a little arbitrary but I think it's important to the paths not feeling railroady. If I, as an experienced player, can just focus on developing my empire well and naturally hit legacy milestones for whatever I'm succeeding in, that feels much more fluid than "You hit 300 culture per turn! Here is culture legacy milestone number 2!", even if 300 culture per turn can ultimately translate into all kinds of other, more generalised strength).
Basically, if we're having this system of ticking off boxes each age, the best implementation I can see is one that serves as a helpful guide for new players to improve, but that experienced players can largely forget about and just naturally complete as they play.
To be clear, I absolutely would like legacy paths to be as open as possible and have more ways to complete each one; these are just my rough thoughts on the good and bad of how they're currently implemented.
