Well I mean just look at Bitshares... that's the first DPOS network @dan created... its sole purpose is to be a DEX, and no one uses it. Rank #500+ on the market cap.
The ideal DEX would be in consensus with every network that it lists, so if you wanted to trade BTC, LTC, XMR, and Hive then anyone running a DEX node would need to run all four nodes at the same time. Then every new listing would require running a new node. That's simply not a feasible strategy.
There are other complications as well, such as escrow. Who controls the money while it's locked in escrow and what happens when there is a problem?
I have no idea how ThorChain works, because actually getting a DEX to work is difficult and complicated. ThorChain liquidity pools have been hacked multiple times over the last few years. They'd be much higher on the MC if they never got hacked. Every liquidity pool with money locked in it is a massive financial incentive for hackers to crack.
This is just a very deceptively complex problem best left to be solved by another network who's sole purpose is to solve it (like thorchain). At the same time, the foundation of a DEX are oracles that act as a proxy for the node consensus problem I described earlier. Because witnesses are indeed trusted oracles we might have the tech to make something happen. But honestly once we get a connection to BTC and maybe EVM there's little incentive to push further unless it becomes a copy/paste solution AND creating new pools doesn't dilute the old ones.