When trying to determine the state for a member of a cluster, it is typical to use /dbms/cluster/status to poll health as described in the following documentation:

- https://neo4j.com/docs/operations-manual/current/clustering/monitoring/endpoints/

When starting a cluster member, there can be a window of time where the member goes from "starting" to "online".  During the time when the member is in "starting", traffic cannot be routed.  However, this endpoint will return with a 200 OK even if the member is still "starting".  Although there is a /ready endpoint for database health, there is not a member/node equivalent.

When using SHOW DATABASES to see the currentStatus for a given database, starting refers to the state of that database on that member and not to the overall health of that server/node/host. There can be several databases hosted on a server in a variety of states, as shown in the following documentation:

- https://neo4j.com/docs/operations-manual/current/database-administration/standard-databases/listing-databases/#database-states

The way to work around this problem involves the JSON array that is returned with /dbms/cluster/status.  In the result array, there is an entry for each database hosted on that particular server, in which the entry for database that is starting will have the key "healthy": false.  It would be sufficient to inspect the result and check all databases are healthy, which would conclude the member itself is also healthy and ready to receive traffic.