Reported from the Google groups:

https://groups.google.com/forum/?fromgroups#!topic/mongodb-user/eXZfE8EprBI

Description from user:

1) restore host.com with latest backup (including oplog) and execute
rs.add("host.com") - and you'll that data will be deleted and resynced
from scratch
2) restore host.com with latest backup (including oplog) and add it
into the replica set with rs.reconfig - it will be catched up as slave
almost immediately
3) try to reconfigure one of shards with rs.add and rs.reconfig - and
check the mongoses logs - you'll see the difference