• We created a new mongo node.
• We added it to an existing replica set as a hidden secondary.
• It went through startup2 and eventually came up as a hidden secondary.
• Clients connecting to it would receive the following warning.

  2015-11-06T17:41:46.790+0000 I STORAGE  [initandlisten] WARNING: the collection 'some_db.fs.chunks' lacks a unique index on _id. This index is needed for replication to function properly
  2015-11-06T17:41:46.790+0000 I STORAGE  [initandlisten]          To fix this, you need to create a unique index on _id. See http://dochub.mongodb.org/core/build-replica-set-indexes
  

Looking at the code, it seemed like it was because the index descriptor object had been called even though its destructor had already run.

Workaround:

• We restarted the mongod process on the hidden secondary and now new connections don't have that warning.