Currently mongod is limited to only asking the local cell's CMS for the ORP address. If the local CMS is down, mongod cannot find an ORP at all even though the other cells CMS are already in its config and any ORP is able to serve the LSNs we want. mongod is expected to be able to tolerate a single CMS cell outage, and the ORP project now puts this into the availability path.

This ticket should ensure that ORP addressing tries each cell in its config preferring local first and falling back to others on failure. mongod already does something similar today to get the segment list from CMS. We should refactor the logic to try each cell into a shared helper.