Short of a mathematically provable algo, my assumption is that full information + clever guys + lots of servers will be able to unravel any path information. From an anonymity standpoint the whole txout->txin system is absolutely horrible. Its like if you looked at a $100 USD bill with a microscope you could see every transaction it was involved in! This is why some theorize bitcoin was invented by one of the TLA's
For example, at some point the funds come out of the stealth address to a public destination. Even if knapsacking analysis is prevented by streams of identical denominations, there is still timing analysis.
That is partly the reason that darkcoin is 99% anonymous. You run into the lock and key problem:
Any security system !ANY! is only designed to defeat a certain level of motivation, an anticipated level of attack. No system can be 100%. In the lock and key analogy, if you really build an impenetrable lock system, an supremely determined attacker will simply attack the door itself. You can build security against a reasonable level of attack, but you cannot protect against any attack. It simply isn't possible.
The goal is to make it more costly to attack something than any potential gain. That's why you can put a chain lock on your beat up bicycle and an expensive alarm on your Ferrari. Darkcoin cannot reasonably expect to defeat the full computing power of the NSA armed with quantum computers backed by picobytes of datastorage to analyze every single transaction within the network and calculate probabilities for in and out addresses. That being said, we will probably never have that kind of scrutiny - in fact, it is likely that most of that power is busy collecting information on Russia, China, etc. While it is theorietically possible for darksend to be cracked, it is highly improbable that it will ever be cracked because the incentive isn't there.
That, and most everyone else is incapable of cracking it.
