After a graceful shutdown, while a cache is missing `num_owners` members, we don't install a topology (and don't start the cache). This behavior can be overridden by utilizing an API to set the topology as stable, with the option to force it. However, it should always be possible to set the current topology as stable for replicated caches without requiring the force flag.

The behavior of cleaning the storage remains. Therefore, replicated caches that don't have enough members will still have the storage cleared. The behavior is as setting the topology with the force flag.