Yep the client could choose many providers to replicate the data, and if one storage provider fail, the protocol automatically choose another to reach the desired goal of replication.

Thus any pleb could find usage for the unused disk space of their always on computers like raspberry Pi etc...

The only obvious flaw that I see here is that it would be difficult to be sure that your replicated data are effectively stored by distinct people, particularly if some big institutional actors enter the market