The way I understand it, and I could be very wrong, is that it's like brute forcing the universe while wearing a blindfold. Because the power of a qubit is exponentially higher than the same number of bits, you can get a lot more information from the same amount of processing power. However, if you measure the qubit, it loses all of that information. Instead, you have to set parameters that say things like: solve for x, and then you wait for the solution to be presented from the qubits. The catch is that you can't see how the qubits are working, because if you do, you observe them and the data is lost. You just have to hope that they solve the problem for you. In reality, it wouldn't be that strange of a process, because you wouldn't ask it theoreticals, you would have it solve complex problems that can be solved in some way. That's why computer gaming gets no benefit, you aren't asking for answer to complex problem, you are telling polygons where to be.
I'm also not an expert on this, but from my limited knowledge, the problem with measurements collapsing a qubit into either 1 or 0 is because they're so small and so finicky that any measurement is by nature destructive. They're small enough that throwing a photon at it causes them to lose information, and you can't measure them without throwing a photon at it.
Again, not an expert, but as far as I know the data exists as some value between 0 and 1 (a float, basically). You just can't tell what it is precisely.
You can control them by blasting them with radio waves. Then, when you're done giving them operations in this manner, you shine light on them and if they fluoresce, it's a 1, if not, it's a 0.
The probability of it being a 1 is dependent on how big the float between 0 and 1 was (so, 0.8 has an 80% chance of being a 1 and a 20% chance of being a 0).
I could be horribly wrong, this is just what I picked up from some baby's first quantum computer explanations I stumbled across.