Slave failover policy

[Michael Kouremetis]

Apologies if asked already (only checked ~250 posts back).

Was just curious on the behavior of Redis Cluster failovers when sets of Master nodes go down. Seems like there are two cases (that I have seen) for when the cluster cannot vote up any Slaves. 

One case is where the majority of Masters go down and thus the cluster cannot vote up any Slaves to take over, leaving the cluster in an uncovered slot state. The Slaves can still be commanded to takeover for the Masters but it must be done by an administrator with CLUSTER FAILOVER command. Will this still be the longstanding policy going forward?  (FYSA: I confirmed this behavior with Redis 5.0.2 on a 6 node cluster (3 Masters, 3 Slaves))

The second case is where all the Masters go down, leaving all Slaves functioning but helpless. I have seen this discussed in (https://github.com/antirez/redis/issues/3697) and then presumably fixed(?) in (https://github.com/antirez/redis/pull/3699). However I tested this with Redis 5.0.2 on a 6 node cluster (3 Masters, 3 Slaves), manually taking down all 3 Masters at once. The Slaves dont seem to do any election and remain Slaves, and thus cluster is in uncovered slot state. Of course you can still do the manual CLUSTER FAILOVER command but nothing is done by the cluster on its own. Is this the intended behavior as well?

Thanks much.