If a cluster is starting up after 'shutdown cluster' it will wait for all the nodes to enable state-transfer and recover from the shutdown completely.

If this is not possible there should be a command to bring the cluster back to AVAILABLE state.

If there are caches with unshared stores and the resulting cluster miss a segment (missing Nodes >= numOwner) there should be a warning like "loss of data for cache  A, C, n - use f" and the option 'f' should force the restart.