Your answer here is to run multiple nodes simultaneously, and upgrade them one-at-a-time. This will give you extra redundancy anyway, which sounds important for your use case.
You can share keys between nodes using dumpprivkey and importprivkey, if that will help you. And you can send API requests that read the blockchain's state to any of the running nodes.
BTW, as a disclaimer, at this stage of its development we cannot recommend using MultiChain in a such a critical situation.