Here is why I think it will not scale.
From the whitepaper
"Files on Sia are distributed
to a large number of randomly selected hosts."
Intuitively this does not make sense to me.
This should not be random distribution if you care about access times. It should be according to an optimization algorithm. Usually it's something like ant colony or something where the chunks (not complete files) are dispersed in such a way so that accessing them becomes faster the more they are requested. Popularity should have some influence on duplication in the algorithm. So the chunks would have to be deterministically distributed to a specific number of hosts, but I will not nick pick too much on that.
Honestly their algorithm is quite exotic. It will probably work but because it's so exotic there will be few people who will relate to it in a way as to write apps for it for instance. The developers of Siacoin will have to post here or on Bitcointalk so we can explore the algorithm further but I do see how it could have problems scaling.
MaidSafe is an orthodox design in a lot of ways. The areas where it tries new things will either be a major success or will fail in the testing phase which is going on now. So we are finding out right now whether or not their design has merit. The rest is just about finding an economic model but I think once you built MaidSafe there will be enough natural inertia from the 500 developers they claim to have working on it to have an economic model emerge naturally if they let it.
Developers will charge Safecoins for their apps and Safecoins will gain value quickly if the apps are there. If the apps are not there then Safecoins may end up being not worth very much. I'm taking the gamble that the apps will be there and bought Safecoins in the presale as a way of getting a lifetime supply of them for the time when there is an app I have to use.
None of us can forecast.