Unfortunately no, due to the limitation of the the current design of Longhorn.
One volume can only have one engine. The engine creates the device on the node and K8s mount the device into the pod.
In your proposal, that would be one pod with one engine and connect to one replica of the same volume.
But Longhorn maintains the consistency by dispatching the IO from engine to all replicas at the same time. Not by syncing the replicas afterward.
That is, in your proposal, all the replicas won't have data consistency.