Personally I think option A would be a good start, depending on how much work B would be it is possible the time cost is not going to be worth the gain. In an ideal world there would be less reliance on public Hive API nodes - but it is far from realistic for every top twenty witness to run their own Hive node to draw blocks from at the moment.
RE: Hive-Engine RPC Failover Is Not Rolling Over Cleanly, Here Are the Two Fix Paths I’d Pitch