They run node specific code with full access to all of my tech, any coin's blockchain, website API, etc. So you can make a tradebot to do automated trading. I chose to use C, well because I am a simple C programmer, and I will be making many library functions.
Since each tradebot is unique to each node there is no assurance of identical results from each node that is required for a global blockchain. However, if tradebot synchronization library calls are added and also some data synchronization then a linked set of tradebots would be able to generate identical blocks of data at regular intervals. This sounds to me like a parallel blockchain or sidechain, not actually sure the proper name.
So, working bottom up, it is quite possible to construct any sort of system you can envision (and code). Any input to the type of library calls you want supported and specific use cases will help me make tradebots that are most useful. With a growing supernetwork, the tradebots will be the best way for talented C coders to make money as I will also make a way for people to obtain popular tradebots and run them to generate profits, which are then shared with the author.
So, some sort of distributed storage should not be so hard once you solve the virtual sector addressing. Yes, you can use the node's public address as the first part of the sector address and the amount of sectors it is contributing as a way to directly address disk space from the tradebot cloud. To maintain availability you will of course need redundancy, but these are things for the higher level file system to solve.