If XDAG was all about code that nobody else doesn't understand, it would mean that the code is most likely not that well written. Imagine Vitalik and his people wrote code that nobody understands. If these guys are good coders they won't only understand the current code base, but they would also be able to fix its flaws assuming that they are interested in putting in the effort.
I'm somewhat familiar with XDAG codebase. Aesthetically, the code is more or less fine (by standards of such altcoins anyway). I'd compare it to very very early nano circa 2014-2015 (even before raiblocks, when it was called mucoin). But Nano clearly has came a long way since then.
I like XDAG very much for it is in quite compact written C and fairly lightweight. But there's also a lot to trash on.
From formal perspective the biggest issue is that XDAG makes some unorthodox assumptions - depending on synchronized real time to determine state transitions; edges of which are fuzzy between peers. It has undefined efficiency and security bounds, which again, is very similiar to Nano. It also suffers from "let's make it DAG for the sake of being DAG".
As it is now, the coin would work just as well with ordinary blockchain algorithm (=heaviest chain, drop orphans). Because in XDAG, once graph tip is selected for the period, all the "weak" parts of the DAG are orphaned. This is precisely what you must not do for DAG to work - the point is to establish pairwise or total ordering, but otherwise you always keep the whole of the graph - in DAGs, you never "cancel" any transaction, ever. Because if you do, you just semantically end up with blockchain, and a more complex one than it needs to be.
Since the original developer dropped the coin, there are basically 2 things that can happen:
1) Big miner or early speculator, or both happens to be a coder or hires one, and starts improving the code (because he has a big stake in it). This has not really happened atm, all changes on github I see are cosmetic.
2) Someone else comes along, forks the code (think what happened to cryptonote).
It's true that clinging to the coin now has no strong economic basis, unless it becomes clear that 1) or 2) is the case.
IMO, improving this coin would amount to a rather drastic hard fork and complete rewrite of how consensus is done (to get the benefit of dag semantics). I'm not sure if its yet worth it to do that, but keep in mind that consensus rules generally can't be "incrementally improved", and this is what really needs fixing in XDAG.
It could be also easier to just start from scratch. Personally I'll wait and see if 1) or 2) is going to happen first.