We already have issues open for the behavior of importing indices when a node is out of the cluster while an index is deleted, however, in the case of shadow replicas on a shared filesystem, if the index is deleted while a node is not part of the cluster, when the node comes back it will try to re-import the index as dangling without any actual data, leading to exceptions like:

[2016-01-14 23:17:23,303][WARN ][indices.cluster ] [data1] [[my-index][2]] marking and sending shard failed due to [failed recovery]
[my-index][[my-index][2]] IndexShardRecoveryException[failed to fetch index version after copying it over]; nested: IndexShardRecoveryException[shard allocated for local recovery (post api), should exist, but doesn't, current files: []]; nested: IndexNotFoundException[no segments* file found in store(default(mmapfs(/tmp/foo/my-index/2/index),niofs(/tmp/foo/my-index/2/index))): files: []];

It would be nice to have an option to disable this dangling import behavior only for indices on a shared filesystem, since this is likely to cause much more confusion than dangling imports on non-shared filesystems.

Related #13298