Given these problems, and the fact that any fixed fee will ultimately not make sense, I still support some kind of proof of model (e.g. PoS) and local/global assets to ultimately fix this problem (as I described
several pages back). It has a natural analogy with how marketing (in the real world) works -- anyone can come up with an idea/product, but to introduce it to the real world, there are costs involved. Even further, it (a PoS function) directly incentivizes holding XCP without introducing a balance requirement (a heavy handed way to do things) or restricting it per address (which = Sybil attack). Again, proof of stake doesn't have to lead to inflation of the underlying asset (one of the most common criticisms); in XCP's case, it can create new assets.
As an extension, I could even imagine XCP holders that don't want to create assets to pull their "proof of stake" power into some sort of pool to help promising/deserving projects that don't have a lot of XCP to launch.