Truly dynamic systems won't allow for the player to be the prime mover (at least, not until fairly late in the game, if that's ultimately what the game's going to be about anyhow). We're not really
butterflies until we're more than noise.
We can talk about the player's relationship to the AI economies (as a predator, consumer, enabler and ally) but there are matters of degree. For example, if Hegemony and Tri-Tachyon decide that this game, they're going to fall into a long-term pattern of warfare with one another, there are big fleets having big battles with noticeable effects on the economies of both sides.
Until the Player's in a fleet that can participate in engagements large enough to be more than chance (both sides are going to lose ships somewhat-randomly during this whole period) they're not all that important. If nothing else, you can see that by playing Exerelin, where a lot of these mechanics are already built- while players can and eventually do play crucial roles, it's not until they're at the point where they can take on multiple big fleets, often without a chance to repair in between. At that point, the player's a real force; before that... well, the player can only be in one place at a time, so their impact is small.
Now, if we're going to ultimately be able to form our own fleet economies and so forth, that makes things a bit more complicated, but not that much, presuming we don't have much if any micro-management capabilities over our economy's fleets, designs and compositions (if we do, well, then we're practically playing another game and the AI issues are in many ways more fundamentally challenging).
Our economic presence will matter, but again, only to the extent that it rises above the noise inherent in these systems. One Outpost with 5 trading fleets in one System is, what, maybe 0.01% of a Hegemony's GDP?
The problem with the idea that the whole "ecosystem" being so "delicate" that player actions "from the start" largely determine what happens is that if it's
that delicate, it'll tip due to noise alone, no player involvement required. Systems like that have to be pretty robust.
In Mount and Blade, this kind of dynamic, yet largely-static system was achieved mainly through their siege mechanics and how the kings and lords were supplied with armies.
Sieges took bloody forever for the AI to execute and failed often. Kings and Lords never really ran out of money (behind the scenes) and got troops overnight, making up for their losses swiftly enough to prevent radical changes from happening.
Ultimately, the important places changed hands very infrequently. It kept things pretty stable and allowed the player, when he/she was finally ready-ish, to launch their takeover of the world.
But the player wasn't a "butterfly"- they were more like a
giant, radioactive ogre covered with barbed wire. They were not noise, essentially; they were making choices that finally mattered and they were equipped to play with these giant robust empires and maybe win.