But how can this "remote-grass-touching-arm" push the "smell" and "tactile" back to the user? Is there an open spec for this? It should certainly be P2P and E2E encrypted. Also "smells" should ensure not to use patented or proprietary names.

Maybe some CSmS, Cascading Smell Sheets? Or TFP, Tactile Feedback Protocol, the one that uses JWT and JSON over HTTP2 and websockets?