For the scenario with a shared cache store, the following request

cache.getAdvancedCache().withFlags(Flag.SKIP_CACHE_LOAD).entrySet()

will return on the local entries and not all the in-memory entries.

The ClusterPublisherManagerImpl assumes it will fetch the entries from the shared store and does not do any RPC to the other nodes. It should check SKIP_CACHE_LOAD and skip this optimization.