Unfortunately, this is a perennial downside with many 4x titles. 4x titles can often be imagined as a race. In the beginning, every racer is racing separately, and doesn't necessarily know who is ahead - which is fun. Often developers have put a lot of focus into the opening moves to make them balanced as well. It's easy to play the first hour and test that, and also those first hours are the ones game reviewers usually see. Plus the whole 'hook them in the first 15 minutes so they don't return it' phenomenon.
However, when the racers come together, most 4x games naturally allow lots of interaction - at which point, the racers in front start tripping the ones behind. Unlike the experience of real-world empires, there is no friction that brings down a 'big empire' eventually. And rubber-banding loser is also a lame-feeling solution. 4x games also usually have multiple, multiplicative places to be dominating in, meaning that small advantages can get more magnified. So an economic advantage turns into a military one and both multiply together against another player.
There are a few solutions to this that have been employed
The "everyone gangs up on the big boy" approach. This can feel artificial and strange, and also means that any sort of diplomatic game is subsumed by the 'real politik' of beating up the biggest threat, no matter who your allies or enemies were in the past.
The "cheat the computer stronger" approach also just usually means the player takes longer to catch up and pass, but the 'bully' problem is not subverted, just delayed - and if that delay is long enough, it allows the AI to be the bully, or demonstrates that the AI doesn't know how to take advantage of its strength
The "multiple win condition" approach. The idea being that somehow you can be winning in one area and losing in the others, and still eke out a come from behind victory. Because being good in one area seems to always make a player good in others in 4x style games, this always ends up being a fallacy - it's usually more of a 'choose your own ending' option for the dominant faction. Or a "I don't feel like hunting down every last unit, I'll just ally with the losers" exit strategy.
The "big empires suck" method. Usually some sort of grit is thrown in the wheels of a big empire, like the happiness mechanic of Civ 5, or the more expensive 'future tech' costs of many Paradox interactive games - which attempts to slow the progress of big empires so little ones catch up. this is just rubber banding in reverse, and usually the design subverts this because big empires are still better. Also, it still needs another solution to allow others to catch up or gang up on you, so it is not in and of itself a complete solution
The "divergent mechanic threat" approach - taken by Gal Civ 2, for instance, with the Dread Lords. Some OTHER thing is your antagonist, isn't necessarily using the 'race' mechanics that you are, and can target the 'bully' civilization without feeling like a complete cheat.
The "really awesome, not cheating AI approach" - never works. Gal Civ 2 is a great example of a really good AI, but even there the players learn to beat it and the AI never catches up. Even if you do have an awesome AI, the player will just end up not playing it, because frankly, who wants to lose the computer over and over because its *smarter* than you.
Out of all of them, I think the divergent mechanic actually works best, and is clearly a good option for Elemental, with what I gather from the back story. I don't think anyone has every managed the multiple win condition thing right (outside of board games - there are some board game 4x style games where that has been done pretty well, like settlers, for instance) but maybe that's another option.
