could the toroid based network protocol be a good infrastructure for replicating information on a network in general, like for example, as an alternative or a competitor to TCP/IP, BitTorrent, or IPFS?
Certainly not all network layers (because you mixed several network layers protocols in your question). In general there are attempts in the wild to use similar approach to replace p2p class protocols. BL protocol is designed to serve BL and give throughput benefits by leveraging machine learning prediction abilities, GPU boost when available (to be close to real time in reaction to changes of net topology) and alike. Thus, it's not meant to be general use protocol, however if it proves reliable and performant (and we design it with simplicity in mind) then with minor changes can be applied elsewhere.