Replication Fails with auto-failover?

couchbase server 2.2, java 1.4 client.

I have a 3 node cluster, and my test bucket has replication.ONE. I kill one of those, and call:
client.set(key,json, ReplicateTo.ONE);

This will fail ~%50 of the time with the following error:
Currently, there is no replica node available for the given replication index (1).

I would have assumed with auto-failover, (that once the failover kicked in), this would not occur?

Am I missing something?

Looking at the admin console, it shows ‘A server was automatically failed over’, In addition it shows the Fail Over Warning: Rebalance Required…

Also what I find weird:
The admin console has the ‘fail over’ button enabled on the node that is ‘down’, and the Remove button is grayed out. (I would have thought it we be reversed).