The Gridcoin blockchain/network should use as little computational resources as possible. Therefore no dedicated servers should be needed. The power of P2P of consumer level hardware should be enough.
Not talking about boinc and the actual scientific computation here.
Scaling to the size of bitcoin or attempt to replace Visa will surely hit a roadblock. However unlikely is that to happen, the problems could perhaps be solved by clustering as described in Freedman&Mazières: "Sloppy hashing and self-organizing clusters".
Rewarding users for their upload, download storage and processing for blockchain, however small it is, would be awesome. But I am not aware of any algorithm that can prove their contribution in a decentralized, consistent and lightweight manner.
I agree with that sentiment. I am running under the assumption that Gridcoin is functional with the amount of processing power we currently have and that it will be sufficient for years to come.
As far as the M. J. Freedman & Mazières idea (link below), it sounds good to me. It also appears to include an algorithm for a emergence capturing mechanism in order to trigger the clusters into self-organizing behavior.
https://www.researchgate.net/publication/2901175_Sloppy_Hashing_and_Self-Organizing_Clusters
Good to know. That might mean that it has not been done yet. Add value to Gridcoin? Maybe we start simple, like just recording how much "Out:" is recorded in the Network Traffic tab of the Debug Console. We can assume that this approach would be fraught with questions. Such as "How do we not record the same network traffic twice?", "What do we do with this information?", and "Do we record downloads too?".