Caches expect to receive an initial state for the other nodes during startup (await-initial-transfer="true" by default).

There is an exception to this rule when rebalancing is suspended: the cache is able to start on a joiner without receiving any state.

However, this exception does not apply when a cache is in DEGRADED mode and rebalancing is implicitly suspended. Instead a joiner will wait until the cache goes back to AVAILABLE mode in order to start the initial state transfer, or fail to start with a generic Initial state transfer timed out for cache %s on %s error message.

We should handle caches in DEGRADED mode the same way we handle caches with rebalancing suspended, and allow joiners to start without waiting.