When a node is created while an earlier lookup is still finishing, the older “not found” response can no longer overwrite the newer state. Reads and updates now recover as soon as the creation is published instead of continuing to report the node as missing until the cache expires or the portal is restarted.